Artificial Intelligence at Human Level by 2029?

Gerard Boyers writes "Some members of the US National Academy of Engineering have predicted that Artificial Intelligence will reach the level of humans in around 20 years. Ray Kurzweil leads the charge: 'We will have both the hardware and the software to achieve human level artificial intelligence with the broad suppleness of human intelligence including our emotional intelligence by 2029. We're already a human machine civilization, we use our technology to expand our physical and mental horizons and this will be a further extension of that. We'll have intelligent nanobots go into our brains through the capillaries and interact directly with our biological neurons.' Mr Kurzweil is one of 18 influential thinkers, and a gentleman we've discussed previously. He was chosen to identify the great technological challenges facing humanity in the 21st century by the US National Academy of Engineering. The experts include Google founder Larry Page and genome pioneer Dr Craig Venter."

Comments

Oblig.

[Anonymous Coward]

I for one welcome our broadly supple, emotionally intelligent overlords.

Re:Oblig.

[Anonymous Coward]

While you're at it, welcome the flying car and Duke Nukem Forever.

Re:Oblig.

[Anonymous Coward]

Exactly--Human-Level AI is the new flying car. Not going to be around anytime soon, certainly not in 2029.

Re:Oblig.

[fyngyrz]

Speaking as an engineer and a (~40-year) programmer:

Odds are extremely good for beyond human AI, given no restrictions on initial and early form factor. I say this because thus far, we've discovered nothing whatsoever that is non-reproducible about the brain's structure and function, all that has to happen here is for that trend to continue; and given that nowhere in nature, at any scale remotely similar to the range that includes particles, cells and animals, have we discovered anything that appears to follow an unknowable set of rules, the odds of finding anything like that in the brain, that is, something we can't simulate or emulate with 100% functional veracity, are just about zero.

Odds are downright terrible for "intelligent nanobots", we might have hardware that can do what a cell can do, that is, hunt for (possibly a series of) chemical cues and latch on to them, then deliver the payload -- perhaps repeatedly in the case of disease-fighting designs -- but putting intelligence into something on the nanoscale is a challenge of an entirely different sort that we have not even begun to move down the road on; if this is to be accomplished, the intelligence won't be "in" the nano bot, it'll be a telepresence for an external unit (and we're nowhere down *that* road, either -- nanoscale sensors and transceivers are the target, we're more at the level of Look, Martha, a GEAR! A Pseudo-Flagellum!)

The problem with hand-waving -- even when you're Ray Kurzweil, whom I respect enormously -- is that one wave out of many can include a technology that never develops, and your whole creation comes crashing down.

I love this discussion. :-)

Re:Oblig.

[Easy2RememberNick]

Wasn't there a simulation of a mouse's brain, or a few cells of it, for a few seconds with the help of a modern supercomputer, we can barley manage to do that.

  Even if an intelligent computer was somehow created it would be an enormous accomplishment to have it be as intelligent as a bug or a small animal. All it would be capable of doing is reacting on instinct (which the programmers would have to create).

  Emotions and language seem very far off, I'd say such a thing is centuries away.

Re:Oblig.

[ILuvRamen]

well as a programmer for 2 years and someone who just knows what he's talking about, computers and human brains will not be equal anytime remotely soon. They work in the most basic way completely differently from each other. We can look at a picture and say it's Steve Jobs in a sombraro and a million computers couldn't figure that out in years of facial recognition and object comparison. And yet it'd take a room full of humans with pencils and paper to figure out certain math operations that computers can do in under a second. Computers are linear and human brains and dynamic and relational. Computers can't do hardly anything open ended and we can't do linear stuff very well. And yes you can write code for relational database access but it doesn't come close and the IO time on any storage media can't come close to a human brain.

Re:Oblig.

[Anonymous Coward]

Odds that we will have human-level AI in the next 50 years are close to zero.

Things like cold fusion, teleportation, quantum computing, virtual reality capable of universe-scale simulation, therapeutic gene engineering and nanosurgery, universal molecular constructors, interstellar flight and perhaps even Dyson spheres... all these we will get before we trully can start getting at building the AI that is human-matching. At least we know how we can handle all the other problems, the advances they require and the research that is still needed in their fields.

But sadly, we still know jack shit about how the brain works, or how to make one in silico (and I am saying that in all seriousness, fully aware of the staggering amount of current knowledge about anatomy, physiology, and psychology of the brain). Consider for instance that brain alone requires over 50% of all of our genes; any other organ (your skin, your penis, heart, kidney, lungs) needs less that 5%. Even trying to put together the very few known protein-protein interactions quickly turns into a giant clusterfuck of data, with degree of complexity growing as a factorial of the number of proteins. If we correctly assume that consciousness is the result of a kind of gestalt of the protein interactions, the anatomical wiring, the multiparallel computation by trillions of cells, and the acquired experiences, building AI is a near impossibility until we get both the revolutionary math tools and the quantum computers capable of universe simulation.

Part of the problem is (to paraphrase Ramsfeld) that we don't even know what we don't know about brain. And no, simply copying the brain structure will not the answer.

Re:Oblig.

[Anonymous Coward]

well as a programmer for 2 years and someone who just knows what he's talking about, computers and human brains will not be equal anytime remotely soon. They work in the most basic way completely differently from each other. We can look at a picture and say it's Steve Jobs in a sombraro and a million computers couldn't figure that out in years of facial recognition and object comparison. And yet it'd take a room full of humans with pencils and paper to figure out certain math operations that computers can do in under a second. Computers are linear and human brains and dynamic and relational. Computers can't do hardly anything open ended and we can't do linear stuff very well. And yes you can write code for relational database access but it doesn't come close and the IO time on any storage media can't come close to a human brain.

Let me guess.... you're volunteering to confuse the "AI vs. Human intelligence" issue by becoming the first human to fail the Turing Test?

Re:Oblig.

[Your.Master]

Let's imagine that computer "processing power" doubles every 2 years for the next 20 years, from a combination of hardware advances and software algorithm innovation. That's not quite Moore's law, and it's not really likely to work smoothly just like that, but just take it as one possibility. In that case, computers of 2029 will be 1024 times as powerful as today. So the question is, are human brains = 1000 times as powerful as a mouse's brain?

Maybe they aren't. But when you say a few centuries, I can't agree anymore. Let's imagine one century. Now we're hitting 1.12589991 × 10^15 times. A human brain is CERTAINLY within that complexity range. The caveat here is can we maintain the doubling rate for a full century? Well...Ray thinks we'll do far better than that (his "law of accelerating returns"), I'm not convinced we'll even be able to sustain the rate -- I think honestly we're looking at a plateau maybe 10, 20 years down the line, and will look back at computing as an S-curve until the next big breakthrough which nobody can predict. In my view the last couple "next big breakthrough"s happened at convenient times to make it look like we weren't following an S curve but we're just getting sharper and sharper, but I don't see any reason why the next one should happen just as conveniently. But since it's unpredictable, I could blindsided by it and it could happen next week.

Language isn't far off at all, we just about have it already. Emotions are nebulous and some people will move the goalposts forever, while some may prematurely be convinced by a video game character. I'm not necessarily convinced they are the hardest part of this. I don't know how to make them, but I don't know how to do the rest of this too. I just often see emotion being listed as the be all and end all most difficult task and I've never seen any reason to believe that to be so.

Re:Oblig.

[fyngyrz]

Wasn't there a simulation of a mouse's brain, or a few cells of it, for a few seconds with the help of a modern supercomputer, we can barley manage to do that.

Well, let's look at the rate of general progress in computing. In 1971, we were putting 2300 transistors on a chip. They ran at a few hundred KHz. In a fairly smooth progression, we've gotten to 3 GHz, where we're likely to stay, and today, we're at about two billion transistors on a chip, with no end in sight as to how far that can go. This is not Moore's law; Moore's law is about how many fit into a particular space; this is about how many can be integrated into a functional unit. That's 36 years. Thirty six years from now, that ability to "simulate a few cells" should grow just in the *normal* scheme of things into an ability to simulate a billion or so cells without any trouble. But there's more to this. Not everything in a cell needs to be simulated; for instance, metabolic processes such as waste generation and removal don't, nor do breakdown, aging, impacts by free radicals, all of that. Part of what needs to be done between here and the goal is streamline the simulation so that it is operating in the zone of mentation and not biological imperatives. I suspect, and yes indeed this is just my opinion, that the simulation will be much easier when we understand just what it is we need to simulate.

This all leaves out the issue of non-simulating intelligence, where the thinking is not patterned after human mechanisms; this could arise from evolutionary software or something along those lines. And of course, one of the reasons that all this is kind of a holy grail anyway, only the first intelligence is difficult; the second... Nth is just a matter of copying a machine state.

As for language, that's solved in the I/o sense -- synthesis and "listening" are both satisfactorily complete. Intelligent discussion can only be expected from an intelligent machine, so that's only as far away as machine intelligence is.

Even if an intelligent computer was somehow created it would be an enormous accomplishment to have it be as intelligent as a bug or a small animal.

Small animals, I'm of the opinion, are a lot more intelligent than most people give them credit for. They just have a different intelligence. I am sure that we will go through the small animal level on the way to our level, and beyond; the thing is, if you can do the one, you can do the other. There's no indication of a significant difference in the wetware, there's just more of it and it is arranged somewhat differently. No reason to expect anything different from hardware designed to do the same job.

Emotions and language seem very far off, I'd say such a thing is centuries away.

Why? Small animals do both. Those aren't even the hard things. The hard things are introspection and self-awareness. Those are the ones we have not even a theory for, today. In any case, your ideas are certainly in with a lot of good company; but not me. I think we're only one discovery - algorithmic in nature - from AI. Self-awareness may turn out to be a property that self-organizes and arises without any special prodding from us; that would be marvelous, not to mention fortuitous, but hardly impossible - again, that's how nature did it.

Here's why I think we're just an algorithm away. If you left a question that absolutely required intelligence on a counter, and went back to pick it up the next day, and the answer was there -- you would agree that an intelligence had answered the question. If a human could answer it in one second, or an AI could answer it in 23 hours, it's still just as intelligent an answer when you pick it up. The point is that speed really isn't the issue. The issue is the process, that is, the algorithm. So it turns out that in terms of speed, number of transistors, etc, that's really not the limiting factor for developing intelligen

Re:Oblig.

[Walt Dismal]

I'll go Ray one better. We will have this before 2029. My company is working to bring to market synthetic intelligences that among other things have feelings, emotion, and mood, and understand human emotion. One may say, why try to do this? Because in order to understand people, a synthetic intelligence must understand these things, which serve real functions in people, and real functions in AIs trying to operate in a people world.

Note that it is not necessary to build 'perfect robots'. People think, and yet they are not perfect. They make mistakes, yet navigate through life. So we do not have to make flawless logic brains. The way people work is that we try to find good if not optimal solutions to problems, but we do not always exhaustively search for the perfect solution. Thus many problems in life can be solved in different ways than you would expect. We do not have to build a machine that finds the optimal solution to a traveling salesman problem in order to make a system that can walk from the kitchen to the front door. It just has to be able to get there reasonably optimally. Also, we do not have to replicate the human brain in order to think much like a human, we merely have to come up with functional systems that can provide similar functions. For instance, the human brain has the amygdala, which can be likened to an interrupt controller for emotional responses. Well, that functionality can be done in a hardware-software system that reasons about priorities of tasks and goals depending on their current 'value' of urgency to the 'brain'.

Many current researchers in many cases are missing the mark. For example, as good as it is, the widely-used AI textbook by Stuart Russell and Peter Norvig (who heads research at Google) has major omissions. It does not dwell on key things needed to bridge between AI and human psychology. Other things like the OCC model of emotion used in AI is incomplete and incorrect in parts. A new approach has been needed, and one I've been developing for decades in stealth mode. I'm writing a 5-volume book set on it. I want it to be the Knuth set of AI.

I'm in the process of patenting the mechanizations of my underlying technologies, and trying to cut deals with companies making multicore processors so their architectures support the thread swapping needed to make virtual neural nets practical. Once we get a 1024 simplified-core processor that supports virtual NNs, it'll be a lot easier to build a machine with many of these that does for NNs what disk swapping does for OSs, than to build a billion-neural processor hardwired machine. And easier to do visual perception systems properly too. So Ray is right. If I can drive certain companies to build the right silicon, we can get there by or before 2029. My current software does what I said, but it's too slow on current hardware. Needs new processors and new system architectures, and it will take 20 years to get the infrastructure all built up. But not a lot more.

Re:Oblig.

[fyngyrz]

But sadly, we still know jack shit about how the brain works

Most of us know jack about the algorithms that allow us to catch a baseball in flight, yet we can still do it. Furthermore, a person from 10000 BC with no math at all by today's standards could do it just as well as we can. Implementing solutions does not always require a complete understanding of what you've done. You can even be wrong and it'll still work for other reasons. So hard-pegging this to what we "know" could be a severe error.

And no, simply copying the brain structure will not the answer.

That's a very bold statement, especially since (a) that's the way nature does it for all its intelligences, high and low, so we know the process works in the general case, and (b) as you say, we don't know many things yet, so claiming that we "know" what won't work seems to be disingenuous or at the very least not well thought out.

I think it is important not to conflate the fact that we don't understand something with the idea that it will be difficult once figured out or discovered as a consequence of some fortuitous sequence of events. That's been shown again and again not to be the case. It *may* be so, but it is by no means certain to be so, and for that matter, it isn't indicated by the complexity of the brain's hardware. The brain is considerably more formidable as a mass of immensely complex moderated connectivity than it is as a collection of cellular-level mystery machines, and a good deal of the complexity at the cellular level is almost certainly irrelevant to the task of thought -- keeping the cell alive is probably in no way related to non-pathological mental operation, yet there's a lot of hardware and systems involved in the task.

Re:Oblig.

[fyngyrz]

My current software does what I said, but it's too slow on current hardware.

That's a *huge* claim; if it is true, you have AI now. Because -- as I explained in a previous post in this thread -- speed is absolutely irrelevant. If you can demonstrate your claim that your software operates now, no matter *how* slowly it operates, you are at the end of your funding issues, not to mention any other issues you may face in life. Which -- to be frank -- is why I doubt your claim. At the point you explicitly claim to be at, I'd already own a mega-yacht and be pulling up next to a lot of potential love.

But good luck, and I really mean that. I'd much rather be wrong and see you bring this right to the table, even if you have completely blown the financial potentials of the development process.

Re:Oblig.

[Kagura]

Wasn't there a simulation of a mouse's brain, or a few cells of it, for a few seconds with the help of a modern supercomputer, we can barley manage to do that.

We already know that it's possible to contain 100% of real-time human brain functions in a casing 10cm by 10cm by 10cm and weighing under five pounds. Now we have to build one from the ground up with potentially slower, yet better understood technology. The problem, unfortunately, isn't related to hardware. I have no doubt that processor power will soon be sufficient for our needs, but without software that can think on the level of a human, it's just another personal platform to play Duke Nukem Forever on.

Re:Oblig.

[DamnStupidElf]

Things like cold fusion, teleportation, quantum computing, virtual reality capable of universe-scale simulation, therapeutic gene engineering and nanosurgery, universal molecular constructors, interstellar flight and perhaps even Dyson spheres... all these we will get before we trully can start getting at building the AI that is human-matching. At least we know how we can handle all the other problems, the advances they require and the research that is still needed in their fields.

On the other hand, making a machine with human intelligence is (literally) as easy as making a baby, and humans are very adept at modifying existing tools. We already have working neural interfaces to simple prosthetics, so it's not a stretch to think that intelligence amplification or augmentation is unobtainable in the next two centuries. As we improve our own intelligence, the creation of AI will become easier. It's likely that the hard problems you've mentioned will actually be solved after humans have improved their brains, since any increase in intelligence will make those problems easier to solve, and we seem to be closer to neural improvement than we are to sustainable fusion (we actually have neural interfaces that work; we still can't sustain a thermonuclear fusion reaction).

Re:Oblig.

[flyingsquid]

I'll go Ray one better. We will have this before 2029.

Yes, but what do they mean by "human level intelligence", in particular, which human are we talking about? I mean, if "human level intelligence" means "as smart as George W. Bush", then I wouldn't trust that machine to handle my taxes, let alone any really critical tasks.

Re:Oblig.

[imipak]

Yeah.

Or to put it more simply: Kurzweil talks bollocks.

Re:Oblig.

[ralphbecket]

Speaking as someone with a PhD in AI, I'm very, very skeptical about having human-level AI by 2029.

Whatever definition of intelligence you choose, it probably includes learning and reasoning components. We have some effective learning algorithms, provided your domain is very specific and you have boat loads of training data. We have next to no good reasoning algorithms. Complete search is a dead duck and incomplete search is not very reliable. Worse, search algorithms get seriously confused when the data base is inconsistent (humans are good at maintaining several incompatible world models simultaneously). And that's all before you consider that we have no psychological models of human reasoning that are anywhere near being specific enough to guide an implementation project (please don't mention "Society of Mind"). Finally, there is precious little funding out there for this kind of research, which is a shame, but there you go.

Re:Oblig.

[stjobe]

Chris McKinstry? Is that you? I thought you were dead?

Re:Oblig.

[dookiesan]

We're too obsessed with mathematics. We spend so much time figuring out clever ways to do integrals, and I just don't believe that humans do this sort of thing when making a decision. There has to be reasoning without probabilities. Since you're a PhD in this field any interesting references on the topic would be appreciated (speaking sincerely).

Re:Oblig.

[khallow]

Things like cold fusion, teleportation, quantum computing, virtual reality capable of universe-scale simulation, therapeutic gene engineering and nanosurgery, universal molecular constructors, interstellar flight and perhaps even Dyson spheres... all these we will get before we trully can start getting at building the AI that is human-matching. At least we know how we can handle all the other problems, the advances they require and the research that is still needed in their fields.

Nonsense. The only items on your above list that we have any idea on are quantum computing and genetic engineering. At least with AI, we have a natural way to create intellgence (kids) and so that gives us something concrete to work with.

Re:Oblig.

[Anonymous Coward]

So, just curious, from your comment then I take it you consider 2029 to be "anytime soon"?

'Cause I, at least, do not.

Re:Oblig.

[Anonymous Coward]

I call bullshit. My laptop's already smarter than most Hollywood celebrities.

Re:Oblig.

[akuzi]

> we've discovered nothing whatsoever that is non-reproducible about the brain's structure and function,

That's probably because we have discovered little about the brain's structure and function.

> all that has to happen here is for that trend to continue

Well the field of AI in the last 40 years has made practically zero progress towards human-like intelligence , but I agree with you - the trend will likely continue.

Re:Oblig.

[chrisG23]

But sadly, we still know jack shit about how the brain works

Most of us know jack about the algorithms that allow us to catch a baseball in flight, yet we can still do it. Furthermore, a person from 10000 BC with no math at all by today's standards could do it just as well as we can.

Apples and orangatuns. The algorithm to catch a ball does not need to be known to catch a ball, what needs to be known is 'how to catch a ball' (whatever "knowing" is). However, to program a computer to simulate thought accurately, an accurate algorithm for thought (or the biological underpinnings of neural activity) IS implicitly required, as algorithms are the way the computer works.

Also, we are studying the brain in all its complexity by using a tool, our brain. I think it is impossible for any one brain to fathom how a brain works completely, just as it is impossible to build an accurate model of the entire universe, as it would be at least as big as the entire universe, and therefore would not fit (in the universe). Maybe with the assistance of computers or other tools an understanding could be achieved, but not by man's natural computer alone.

Re:Oblig.

[dubl-u]

The good news is that Kurzweil put cash money down to back his opinion. In this case, Mitch Kapor (of Lotus and OSAF fame) is betting against him.

Re:Oblig.

[themacks]

...we've discovered nothing whatsoever that is non-reproducible about the brain's structure and function...
Seeing as consciousness is considered a function of the brain, I would say that we most definitely do not have the capability of reproducing it, yet alone understanding it. For a working AI model of the human brain, neuroscience and psychology will have to come a long way.

Re:Oblig.

[fyngyrz]

However, to program a computer to simulate thought accurately, an accurate algorithm for thought (or the biological underpinnings of neural activity) IS implicitly required, as algorithms are the way the computer works.

No, you have missed my point. An algorithm or algorithms is certainly required, and I never meant to imply otherwise. Human understanding of said algorithm(s), however, is explicitly not required. And there are many paths that lead to such a situation. Whether one of those will take us to a form of AI remains to be seen, which is what I was saying.

I think it is impossible for any one brain to fathom how a brain works completely

It is one thing to understand the mechanism required for operation -- it is quite another to understand the state it is in. I think you are confusing the latter with the former; the former is relatively trivial, and the latter is not required any more than a complete understanding of the state of everything involved at NASA is required in order to create, launch and recover the space shuttle. Complex systems are holistic, mostly co-operative combinations of subsystems, and as long as someone, somewhere, understands (or understood at one time, or possessed an adequate analogy to, or approximation of) the subsystems, or even the subsystems that make up the subsystems, that's sufficient to develop a fully functional macro system. And -- most importantly -- it only has to be done once, because of the unusual copyable nature of the result.

Re:Oblig.

[hedwards]

Odds are extremely good for beyond human AI, given no restrictions on initial and early form factor. I say this because thus far, we've discovered nothing whatsoever that is non-reproducible about the brain's structure and function, all that has to happen here is for that trend to continue; and given that nowhere in nature, at any scale remotely similar to the range that includes particles, cells and animals, have we discovered anything that appears to follow an unknowable set of rules, the odds of finding anything like that in the brain, that is, something we can't simulate or emulate with 100% functional veracity, are just about zero.

Sure if you build in enough memory and processing power the bottle neck ends up being the designers, but it'll be a really long time before the hardware gets to the point where that's possible.

At present hardware will crash if a few bits get in the wrong places or if they're stored incorrectly, one of the things about organic lifeforms is that our consciousness doesn't cease to exist if one of our neurons misfires, at worst we get a seizure or possibly a hallucination. Any machine that's going to surpass us would have to turn those wrong bits into something meaningful without human intervention. Even if they are just unexpected rather than outright wrong.

It may very well be that computer technology will solve that problem, but quite a bit of what we are comes from these random misfirings and unpredictable unreliable results. Modeling what humans are presently like, or even modeling what humans are like at the point when this becomes realistic is far easier than creating something that will outdo us by intellect.

I'm somewhat skeptical when you say that nothing a person's brain can do which cannot be modeled by software, when it comes to talking, moving building, following instructions and things of that nature, I see no reason why a machine couldn't be taught to do those things as well as we do. But when it comes to more subtle things, things which require creativity, sometimes things which require for a deliberate violation of typical common sense, I'm skeptical that a machine could be taught to do so.

I'm especially skeptical about that considering that we don't even know most of the things which the human brain does or how it does it. We know many things, and we know enough to greatly benefit ourselves, but there are still a fair number of things which we don't understand about the brain. It is not a simple organ to understand, just in the last 10 years the amount of information gained about it is sufficient for me to suggest that you shouldn't suggest that there isn't a part of the brain which cannot be simulated.

I really don't want to suggest that it is impossible for us to create something that surpasses our own selves, but doing so would require things which we haven't even dreamed up yet, teaching a computer AI to be capable of meaningful creativity isn't something which is yet even on the most distant horizon, none of the programming languages or tool kits that are available presently offer that sort of capability in anything which resembles a reasonable number of lines of code.

Re:Oblig.

[fyngyrz]

Complete search is a dead duck and incomplete search is not very reliable.

...in humans, specifically. So in the goal of creating a human-like intelligence, we're already there.

Finally, there is precious little funding out there for this kind of research, which is a shame, but there you go.

There are 24 hours in a day, normal jobs take 8 or so, we sleep for 8 or so, weekends offer 32 more hours, that adds up to a *lot* of time for pursuing the tasks at hand without requiring any funding at all, presuming you are motivated and own a decent computer (I just bought an 8-core machine for under three grand, I wouldn't call that a "funding problem" for anyone with even a lightweight drive towards working on these problems.) This is -- I think -- an algorithm problem, not a hardware problem, which means that what you need is a language you can write in (free) and create powerful output (free) and you also might need to collaborate with others, which is also available for no more than the cost of a network connection and some additional free software. A terabyte of long term storage (HD) is about $200 right now, and getting less expensive all the time. Ergo, no funding problem. Unless I'm wrong, and hardware that cannot be simulated is required, which I *really*, *really*, *really* doubt, as nature has never presented us with such a case as yet. But I'll allow it as a possibility, because we're not there yet (though I'm willing to bet money against it... :-)

Re:Oblig.

[fyngyrz]

That's probably because we have discovered little about the brain's structure and function.

No, it is *probably not*. It *may* be, but since *nothing else* has presented us with that kind of problem, the odds of the brain doing so are pretty darned slim. You are postulating a heretofore never-achieved discovery in the course of determining how a mundane (by every indication) biological system, constrained as far as we know by the same physics and chemistry everything else is, operates. Considering the *fact* that there is no indication for such a discovery, I'd say you are way out on a creaky limb and you should be asking yourself how you got there.

Re:Oblig.

[fyngyrz]

If by "working model" you mean a set of ultimate specifications, I'm not sure we need them at all. We may, but then again, perhaps not. Lots of tasks get solved in small steps and the power and value of the sum of the solutions is considerably more than one would anticipate. If by "working model" you mean the first actual AI, in the simulated flesh, neuro-science and psychology are probably entirely irrelevant. After all, nature didn't need them to come up with us. And we have some very powerful tools to make up for our relative impatience as compared to the timescales that process operates upon.

Re:Oblig.

[StarfishOne]

The Dyson sphere will be the stepping stone for the Matrioshka Brain! :)

A more detailed article discussing than Wikipedia that is discussing this subject can be found right here.

Re:Oblig.

[Prof.Phreak]

I thought that AI would take care of itself, as well as the rest... ie: we build a machine that can learn, and it becomes super-human intelligent all on its own, and then goes ahead and builds ``things like cold fusion, teleportation, quantum computing, virtual reality capable of universe-scale simulation, therapeutic gene engineering and nanosurgery, universal molecular constructors, interstellar flight and perhaps even Dyson spheres...''

So our job is to make that initial leap to start the process---which might just happen within 20 years.

Also, we don't need to understand or know how the brain works. the AI just has to work, who cares -how- it works. let it figure out how to do it.

Re:Oblig.

[fyngyrz]

Sure if you build in enough memory and processing power the bottle neck ends up being the designers

No -- we definitely don't know this; you're simply making an unwarranted assumption. Until the actual algorithm is known, we have no idea how much memory and processing power will be required. Remember: Time is not the problem; it never was. The claim that processing power is a limiting factor is entirely specious. An intelligent answer achieved in a hundred days instead of a second is no less intelligent for being eight million times slower to arrive.

doing so would require things which we haven't even dreamed up yet, teaching a computer AI to be capable of meaningful creativity isn't something which is yet even on the most distant horizon

Doing *what*??? Honestly, you could have knocked me over with a feather when I read that. Do you teach a child to be capable of "meaningful creativity"? Of course not; it is a direct and natural consequence of intelligence. You won't have to teach an AI any such thing. It'll almost certainly arise all by itself, just as it does with us, and for the same reasons. I'd say it is completely fair to say that if it doesn't, the thing isn't intelligent in the first place.

...none of the programming languages or tool kits that are available presently offer that sort of capability in anything which resembles a reasonable number of lines of code.

No. Look: We *don't know* what the algorithm(s) is or are; so we *don't know* what the requirements are. You can't proclaim our tools insufficient to the task until you know what the task *is*. Add to that the fact that our experience thus far is that even the most complex systems can be 100% emulated - more slowly, but no less effectively in terms of achieving exactly correct function - by the most simple systems just a few bits wide, and the floor falls completely out from underneath your assertion.

...things which require for a deliberate violation of typical common sense, I'm skeptical that a machine could be taught to do so.

Well, again, are we taught to do so? Or do we learn that some situations call for such a response? I don't recall being taught any such thing, personally. If the thing is intelligent, it'll develop this, or something better (and I strongly suspect that common sense is overrated as a final metric of worth in the first place, so I don't see that as even slightly unlikely.)

I'm somewhat skeptical when you say that nothing a person's brain can do which cannot be modeled by software

I'm talking about hardware function - neurons, connectivity, signals. My position is that what the brain does, or what *a* brain does, is built upon the hardware capability it has. My assertion is that we can (following one path) build that capability. Once it is there, and functioning, we're not talking about models anymore. We're talking about it functioning like you because it *is* like you. That won't require any action on our part, or at least, nothing more than a less strenuous analog of parenting.

Re:Oblig.

[harry666t]

And there's still one question, which greatly adds to the complexity of the whole thing.

What if we, humans, do have a soul?

Atheism is also a religion, because you have to believe that there is no God. There's no proof of either existence or non-existence of a supernatural being.

Re:Oblig.

[KDR_11k]

That is if you give it enough time. I doubt building an even more intelligent AI is a trivial task even for an AI that's more intelligent than a human, especially creative steps don't seem like something that can be easily forced to me. The AI would still have to brainstorm and hope to get an idea that it could develop.

Re:Oblig.

[Znork]

But sadly, we still know jack shit about how the brain works,

We know exactly how the brain works. Its one huge neural network. We've made many silicon versions of those; they've found use in everything from spam filtering to mail sorting.

Replicating the brain as a neural network isn't exactly trivial, but there's nothing 'unknown' about it, just lots and lots of work. It also has the slight problem that if you want an intelligence even remotely similar to what we'd recognize as a human one you'd probably have to replicate most of the learning input drives; hunger, pain, pleasure, survival instincts, etc. Neural networks model the world around them in relation to the inputs that drive the most fundamental 'right/wrong' experiences, vary them and you'd probably end up with something entirely different.

Of course, once you finally created a human-like intelligence with survival instincts you'd probably find it rapidly realizing that with humans around its odds of survival aren't that high (particularly as long as we haven't dealt with the ethical issues). And you'd get SkyNet. So.

In the end, it's dubious that we'd actually want a human-like AI, nevermind a super-human one. Idiot-savants are much more useful for practical applications and it lets us have reasonably smart slave type devices without the ethical (and competetive) problems of full-blown sentience.

Re:Oblig.

[brainwash]

I was thinking exactly of this before I read "I, Robot". It suggested there that a primitive robot brain was built - kind of like the mouse brain we have. Then by extrapolation that 'brain' helped to develop further iterations of itself. I really think that's the only way to advance a lot and it's doable.
Human consciousness will not be developed in the near future, since we have no rules for it, but a mathematical/analytical mind could, once we implement lateral thinking. This could then be used to solve the century's problems.
But then again, huge leaps only happen at war time or with combined concentrated effort of the nations. In this capitalistic world I don't see 2nd item happening very soon.

Re:Oblig.

[Knuckles]

So, just curious, from your comment then I take it you consider 2029 to be "anytime soon"?

You must be very young, then. Decades won't look as long anymore once you loved a few of them.

Re:Oblig.

[Knuckles]

And lived them, even.

Re:Oblig.

[John Betonschaar]

If they want to hit 2029 they not only need full understanding and perfect simulation of our brains, but they also need to start raising this 'human intelligence' pretty fast. If it takes 18 years to raise a human child to be able to interpret, value and effect the world around it in a sensible way, there is no reason to assume a computer can do it any faster.

There's a lot more involved in creating human AI, most importantly semantics for all the things around us. And not 1 single set of semantics, but loads of them for loads of concepts. The meaning of anything said or written is so much influenced by culture, religion, emotional state, education, tuition and other stuff like that. It's naive to think you can reproduce human intelligence by just simulating the medium (brains), and even that is miles beyond our technical capabilities and knowledge of the way our brains work *and* develop when we mature.

So in short, this prediction is BS.

Re:Oblig.

[Plutonite]

Most of us know jack about the algorithms that allow us to catch a baseball in flight, yet we can still do it. I don't think that is a relevant analogy. If you don't know how the biological counterpart works then there is little hope for you to replicate it in logical abstractions/hardware. To sum up your statement: our brain still works even though we don't know how it works... now how does that help us in any way? Unless you can copy the functionality (i.e reproduce sentient matter) without knowing what you're doing, then you are stuck with the traditional approach of actually knowing what the hell is being modeled. Everything in AI today - all the neural and bayesian/statistical stuff, is strongly modeled down to the mathematical level. Not only do we we know what we can do but we can *prove* it mathematically (e.g you can prove what set of functions a perceptron can learn).

Which (I believe) introduces problems from Godel..etc in the brain's case (which is sufficiently complex and self-referential). Even if we do copy the brain's functionality, we may never be able to prove it. So we may in fact have to go with your we-did-it-but-darn-if-i-know-how methodology.

Re:Oblig.

[Anonymous Coward]

We already know that it's possible to contain 100% of real-time human brain functions in a casing 10cm by 10cm by 10cm and weighing under five pounds. Do you have a source on this by any chance? I'd be interested in reading how we realized that this is possible.

Re:Oblig.

[unapersson]

"Atheism is also a religion, because you have to believe that there is no God. There's no proof of either existence or non-existence of a supernatural being."

No it's not, it's a lack of theism. Many religious people seem to find it really difficult to get their head around. Religion and gods have absolutely nothing to do with our lives. We don't sit down every morning and pray to the void. We simply accept reality for what it is and don't see anything in our every day lives that needs a special explanation.

I don't know why it's so difficult to understand. It's not much different to being disinterested in football. You'll have groups of people that are obsessed by it and cannot understand how its not a part of someone elses life.

Surely there's something that's completely irrelevant to your life? Tiddlywinks, Tabletop roleplaying games. As far as most atheists are concerned, religion is just another interest, and only relevant when it tries to force itself on our lives. Imagine how cross you'd get if tiddlywinks players got together and tried to force an hour of daily tiddlywinks playing into national school curriculums.

Re:Oblig.

[Anonymous Coward]

I see a lot of big words...and no understanding of any of it.

Re:Oblig.

[MrHanky]

Odds? If only these AI dweebs would pay off their debts for the bets they made for the exact same thing 30 years ago and 20 years ago. Human level AI is always "20 years in the future". But the only AI we have now is one that can process simple information that it has gathered from a strict set of rules (and then it's already much better than we are). Wake me up when an AI knows what it's like to be hungry (a fairly basic experience needed for understanding people, and an AI that doesn't isn't "human-level"), an AI that can cook creatively, and knows whether what it has cooked looks and tastes good, and then can discuss the latest Star Wars remake without resorting to autist level idiocy about Han Solo shooting first or not while you're eating. An AI may beat you at a game of chess, but it won't brag about it afterwards in metaphors taken from a movie or a book of its own choice. Sorry, but it will never happen, not only because it's impossible, but also because there's no need for such an AI. AIs are developed to perform specific operations, not to be your robot friend.

Also, what you're saying about scientific discovery is a tautology: when it "discovers" something irreproducible, it's not yet considered a discovery at all. When something is discovered, it's also discovered reproducibly. That's a fucking necessity for scientific discovery right there, of course science won't find anything else! Outside the scope of science lies all subjective experience, which is where actual human intelligence resides (the idea of intersubjectivity is what makes us recognise understanding, intelligence, in the other person). Your argument seems radically irrelevant.

Re:Oblig.

[umghhh]

Every time I see these arguments I wonder where are the facts and where is the purpose. Just to systematize this:
1. this doubling argument is a nice one - do you know what M.Twain said about extrapolating? see here for guidance:
http://mrkwr.wordpress.com/2006/11/08/mark-twain-on-the-perils-of-extrapolation/
2. what is the purpose of such exercise? I see none. Simulating more gullible, aggressive and plainly stupid indihviduals of the sort that we have already is just plain pointless. Generating an intelligence of some other kind - well how do you recognize it is an intelligence - when it fires nuclear weapons at you? Or when it looks like governor of California?

Re:Oblig.

[nospam007]

No it's not, it's a lack of theism. Many religious people seem to find it really difficult to get their head around. Religion and gods have absolutely nothing to do with our lives. We don't sit down every morning and pray to the void. We simply accept reality for what it is and don't see anything in our every day lives that needs a special explanation.

Amen, brother, couldn't have said it better myself.

Re:Oblig.

[nospam007]

Speaking as someone with a PhD in AI, I'm very, very skeptical about having human-level AI by 2029.

Whatever definition of intelligence you choose, it probably includes learning and reasoning components.

But it could be good enough to serve as POTUS perhaps.

Re:Oblig.

[Nefarious Wheel]

Fractal structures and neural networks are fun, if I remember my Minsky at all. The secret we need to uncover is how to get nanobots to communicate along synaptic lines without dissipating the signal beyond one or two atomata. I think it's these communications and feedback structures that are key to the idea of layered abstraction of neural networks that provide the summarising signals necessary to work toward artificial consciousness. (Or I could be entirely full of snot, depending on the associativity index of this last thought.)

Re:Oblig.

[Nefarious Wheel]

Oh and I'm an engineer and a programmer, too =). But you have two years on me (/salute)

Re:Oblig.

[Burnhard]

we've discovered nothing whatsoever that is non-reproducible about the brain's structure and function

I'm sorry, but this is such a poorly constructed thesis that I have to respond. I studied AI at University some 10 years ago and there was absolutely NOTHING in the area that one would suppose could "produce" intelligence just by scaling it up or via. simple modelling/simulation. Sure Neural Netoworks have some good pattern classification properties but I am personally in the camp that thinks the real problem isn't reproducibility, it's non-computability. I'm with Penrose and, to an extent, Hameroff. It seems every decade someone comes along and says, "only 20 years to go". I will believe it when I see it.

Re:Oblig.

[Burnhard]

"The brain is considerably more formidable as a mass of immensely complex moderated connectivity than it is as a collection of cellular-level mystery machines, and a good deal of the complexity at the cellular level is almost certainly irrelevant to the task of thought -- keeping the cell alive is probably in no way related to non-pathological mental operation, yet there's a lot of hardware and systems involved in the task."

You (and most proponents of AI) have failed to answer any of the philosophical/metaphysical questions one inevitably becomes confronted with, by using the analogy of the brain as "software" and stating that the hardware is irrelevant. I suspect there are cellular-level mysteries yet to be discovered, including possibly quantum action at a low level, that would have a strong influence on the facts of the matter here. It is a rather simple-minded and arrogant "faith" that leads you to believe we have anywhere near a good understanding of how the brain works.

Re:Oblig.

[Walt Dismal]

It's okay to doubt it; until I demo pieces to people, they don't believe it either. I do have strong AI right now running on a Von Neuman class processor, but AI really needs a massively parallel architecture. Speed IS relevant. You can't just run big wide NNs simulated in a Von Neuman, it takes forever. The best architecture I find is where processing and recognition are the same as memory. That is, once the AI learns something, the pattern recognizer also serves as the memory and the belief logic. In a way, the NNs ARE the data storage, there is no separate RAM. And my AI is implemented in a connectionist architecture that embodies symbolic processing in a new way. The knowledge storage is synonymous with the logic. But this requires massive though simple logic units and parallelism. Further, it is not completely self-organizing from the ground up, I view that as a hopeless approach.

I define consciousness and awareness within a pre-determined architecture, not entirely self-organizing from scratch. The visual front-end in particular is very rigid, but I think it is okay because there is little need for an organism to self-evolve completely new architectures, but rather to be able to run cascaded pattern-recognizers. See the work of Biederman for examples of this. The VFE feeds deeper processing doing cognition, and there is feedback from that to the VFE for training. Just as a baby learns to see, and recognize shapes, and build up from that. The front end is trained as the cognitive end learns and grows too.

AI does not spontaneously rise alone from massive databases, either. I view that approach as useless and a false trail. However, human intelligence does depend on belief systems and knowledge, and those continually grow as we mature from infancy. But to create the equivalent of an 18 year old, you have to have what amounts to 18 years of accumulation of knowledge about the world, and draw upon that. And there is a key but proprietary subtlety about that I'm devoting an entire volume to, that is the key to humanlike AI. That volume is essentially a doctoral thesis about consciousness reworked for use by a design staff. As for funding, no yachts yet. But I'm real, sane, and not a charlatan, and have explained my technology to my patent attorney. I expect to be hiring staff within two years. I posted on Slashdot not for glory but to counter all the nay-sayers who haven't a clue what is achievable.

Re:Oblig.

[Fweeky]

At present hardware will crash if a few bits get in the wrong places or if they're stored incorrectly No, generally it's software which crashes or wildly misbehaves when that happens, unless something critical gets put in an incorrect state or error detection kicks in and *forces* a crash.

There are plenty of algorithms which can cope with flipped bits; for example, one of my favourites is the Bloom Filter; bloom filters let you store whether or not you've seen a given item before, in very little space. They're probablistic; they have a chance of thinking they've seen something before when they really haven't, which goes up as they get full.

Hardware faults resulting in bits getting flipped simply makes them less reliable; they might give a false negitive, or more false positives, but provided the error probabilities are roughly within a known bound they can still be very useful.

You could certainly imagine a data structure for a neural network which is similarly fail-soft; a flipped bit changes the connectivity of a single neuron, or the weight it gives to a certain input, but it doesn't invalidate the entire network.

Re:Oblig.

[F452]

I assumed he was talking about the human brain.

Re:Oblig.

[Teancum]

Speaking here as a full-time software engineer who has been involved with the computer industry since 1977 (I guess not quite the 40 years you got.... but pretty close) and somebody who has studied artificial intelligence extensively, including reading most of the current literature, and having even implemented many of the algorithms related to current artificial intelligence research I can say one thing about what you just wrote:

You are completely clueless about the topic and this is utter bullshit!

I hate, I repeat, I hate, I hate, and I repeat one more time, I hate it when supposedly knowledgeable individuals like yourself try to pass off predictions in a field that has so often made predictions of achieving human-like intelligence for so long that normally I drown out such predictions and pure drivel and something akin to anti-gravity research and faster-than-light communications. Essentially something written by an individual who has not really read the literature, doesn't really understand the breadth and domain of the problem, and hasn't even a clue about the steps necessary to achieve even reasonable levels of intelligence.

I'm not saying here that the study of Artificial Intelligence isn't a useless field, and that some very remarkable tools have been developed in this quest. There have also been some incredible advances in the study of human intelligence, and even so much as questioning what intelligence really is. It used to mean a highly evolved sense of intelligence to be able to recite back volumes of information or to be able to calculate "massive" equations "in your head" quickly. I guess on this standard, computers have already achieved a huge level of intelligence... and did so in the 1940's I might add here with ENIAC. Even as late as the early 1940s, it was considered an honorable profession to be called a "computer"... meaning somebody who spent the day with paper and pencil making thousands of calculations each day... usually in the generation of mathematical tables like logarithm and sine tables, but sometimes other tasks as well.

What we know now is that genuine intelligence is far more complicated, and involves emotions and other states of being that simple haven't been achieved. While I have no doubt that eventually somebody will build something like A.L.I.C.E. that will pass the "Turing Test" for Artificial Intelligence in the next few years on a cursory level, it will only help to push back the frontiers of our knowledge of what intelligence really is and only lead us to a deeper understanding of the true complexities of human intelligence really are.

What my real pet peeve here is for somebody like yourself to spout off that true Artificial Intelligence "is just around the corner" and about to come true, and somebody who is not well versed in the field like some journalist starts to believe you. Or you start to post something like this where you claim knowledge about the topic when you really don't.

My own prediction: Fifty years from now, we will still be having "knowledgeable individuals" claiming that genuine artificial intelligence is only "20 years away". Just as it was predicted 30 years ago. I don't expect this to happen in my lifetime, or the lifetime of any of my grandchildren. Human civilization on Mars will happen first.

Re:Oblig.

[sydneyfong]

> And there are many paths that lead to such a situation

For example? Darwinist evolving programs? Well that could theoretically work, but given that our brains had billions of years where we'd expect some results within decades... I don't think the odds are good for this one. Any other "paths" that you could think of?

Re:Oblig.

[Cyberax]

Yes, that's about right. Human brain contains about 1.5kg of brain matter (and it's estimated that around 700g is necessary for high-order brain functions), and mouse brain contains about 1g.

I have made the same estimation personally few years ago :)

Re:Oblig.

[sydneyfong]

A practical note. "No funding" means the AI researchers get paid to do something else. And according to the horror stories of being an academic, there are pressures to perform teaching duties, pressure to get some "stuff" published, etc. Which means there would be little time left to work on this "grand AI scheme"/singularity/whatever thing.

Time is probably not the main bottleneck, it's the mental fatigue that holds things up. I don't know about you but after a days of stressful work I'd rather be at home relaxing with family/friends than to jump onto another mentally demanding project -- without pay!

Re:Oblig.

[greyhueofdoubt]

This is OT, but I'm curious on a personal level what it's like to be an AI researcher. I work in aerospace, myself- I can do my job all day on real things that go really fast. There may be faster things someday, but I feel like I'm on the flat part of the curve.

I think I would have trouble staying motivated in a field like aerospace if we didn't even have airplanes yet. It would be like going to work and building a wing, building a canopy, building landing gear, and just waiting and waiting for that breakthrough: "I seriously doubt we'll have bird-level flight within 30 years."

I realize that modern AI has produced tremendous rewards so far, even if they are below most people's radar. I'm glad that you chose the program you did. I guess I'm wondering if this 'human-level' AI is even something you/others are researching, or if it's something that you believe will come about when enough collateral research has been done- the 'sum is greater than the whole' approach, if you will.

Best,
-b

Re:Oblig.

[oddtom]

Although unapersson gave an accurate response to your statement, I thought I might elaborate one thing concerning types of atheism. Namely, there are two: positive and negative atheism. Positive atheism states that because we have no empirical or logical evidence for a deity, that deity does not exist. Negative atheism states that because we have no empirical or logical evidence for a deity, we have no reason to believe at this time and place that such a deity exists. The former is a positive belief, the latter is an absence of a positive belief.

What you describe is positive atheism and does, in some cases, suffer from the problem that it assumes lack of knowledge implies non-existence. Some positive atheists might argue that because many forms of theism state that since such a deity is supernatural, one could not use natural means to discern its existence, and hence is no better than if it did not exist at all: we would be none the wiser either way from an empirical standpoint.

But you have a point. Positive atheism can often border rather close to religious thinking with some people. If you'd like, "Atheism: The Case Against God" by George H. Smith is a pretty good lay-person's primer on the topic. (Even though he believes negative atheism is just a form of agnosticism.) It's pretty useful for theists who want to think more critically about their beliefs.

Regards, Tom

Re:Oblig.

[drinkypoo]

Odds are extremely good for beyond human AI, given no restrictions on initial and early form factor. I say this because thus far, we've discovered nothing whatsoever that is non-reproducible about the brain's structure and function

So far, we have not discovered the mechanism behind memory, one of the most important functions of the brain as every other activity the brain engages in invokes memory and is affected by it.

So while so far we've discovered nothing we cannot theoretically reproduce, we also know very little about the way the brain actually works. So far about all we know is that it DOES work, and which major parts of the brain are primarily responsible for which major functions, and approximate effects of chemicals which the brain produces (but we clearly do not understand the full implications, or people on prozac wouldn't commit suicide left and right.

Odds are good that nothing is utterly unreproducible - but there are honestly no indications of WHEN, if ever, we will produce human-level (or human-superior) AI.

Re:Oblig.

[Teancum]

No it's not, it's a lack of theism. Many religious people seem to find it really difficult to get their head around. Religion and gods have absolutely nothing to do with our lives. We don't sit down every morning and pray to the void. We simply accept reality for what it is and don't see anything in our every day lives that needs a special explanation.

I know this is very much off topic from the main point of the story, but I can't let this stay here unanswered.

I realize that those who are atheistic in nature don't have religion and gods in their lives, and it is reasonable to presume that somebody without "faith" in a higher being of any kind can still have guiding morals that govern their life.

Still, I have to argue here that there still is a "religion" of atheism, complete with "prophets/oracles", "priests", "congregations", "sacred literature" and other trappings of religion. That it takes other forms and is usually not so formally organized may be true, and even defining "orthodox atheism" can be a bit of a struggle, but all of these do exist. It certainly takes on a philosophical niche that often takes the place of other religious philosophies in terms of guiding principles in your life.

There are also multiple forms of atheism, ranging from "environmentalists" (devotion to environmental causes as a religion), "universalists" (that somehow the whole universe will make sense ultimately), "scientists" (a solid belief that science alone can solve life's problems), "anti-theologists" (opposing any form of organized religion of any kind), and many others. It is very difficult to take such an emotionally charged term like atheism and force any sort of hard stereotypes. But I do argue that you can identify atheism as a religion, including its establishment as a state religion in many cases, and concerns about how it has entered into public institutions forcing out other philosophical viewpoints.

To tie this back to artificial life/intelligence research, I do believe (there is that word somehow showing up) that some sort of religious philosophy will eventually show up in terms of identifying and working with a "soul". I'm defining that "soul" to be the consciousness or intelligence that has an independently operable sphere of influence that can relate with other intelligences, including at a human level of interaction. This isn't to say that such a "soul" can't have a purely scientific explanation either. I also believe that once such artificial intelligences are developed that it will have a tremendous impact on human religion in a large number of ways, including everything ranging from human fealty and devotion to an AI lifeform approaching god-like standing, modification of theological doctrines, to even a view that working with AI lifeforms is a type of blasphemy that should be rooted out of human society (and the source of future wars).

Re:Oblig.

[drinkypoo]

And no, simply copying the brain structure will not the answer. That's a very bold statement, especially since (a) that's the way nature does it for all its intelligences, high and low, so we know the process works in the general case, and (b) as you say, we don't know many things yet, so claiming that we "know" what won't work seems to be disingenuous or at the very least not well thought out.

Copying the structure of the brain in all particulars would produce a brain and so we would still have failed at producing AI. At best we could claim is to have copied natural intelligence artificially.

I agree with your claim of disingenuity or at least a lack of forethought, though. In theory, if we understood the brain better, we would have a better understanding of the problem. We are currently finding quantum mechanisms for various things in our body (including smell and hearing!) and we still haven't identified any mechanism complex enough to account for the process of memory. If that turns out to be a quantum function it might very well be nonreproducible, at least in a familiar form.

Re:Oblig.

[drinkypoo]

On the other hand, making a machine with human intelligence is (literally) as easy as making a baby

You need to be made to understand that we don't really "make" babies. All we really do is supply the raw materials to our prebuilt baby-making equipment and let them do the work. While we can currently observe pretty much the entire process (and observing the first part of the process is in fact one of the major drivers of the internet) we still can't mimic it. Get back to me when we can make a baby without using sperm, ovum, or womb.

Re:Oblig.

[vhogemann]

But AFAIK they where simulating the BIOLOGICAL behavior of the said mouse brain.

Once we fully understand what makes the neural network of a mammal brain work, we'll be able to create hardware and software that is adequate to work in the same way, without the overhead of simulating every biological process performed.

Also, I suspect that the basic processes are behind every brain, and to be more or less intelligent is a matter of scale. Bigger brain, more complex neural networks, more intelligence. So, once we have a mouse-brain-on-a-chip working, it will be a matter of wiring some together to reach the same processing power of the human brain.

Re:Oblig.

[jdoeii]

Actually, Eugene Izhikevich run a successful simulation of the entire human brain in 2005:
http://vesicle.nsi.edu/users/izhikevich/human_brain_simulation/Blue_Brain.htm#Simulation%20of%20Large-Scale%20Brain%20Models
Simulation of 1 second took 50 days on a cluster of 27 PCs (~4.3M times slower than realtime). Eugene is a pretty smart guy, except not as prominent as Kurzweil. Here is Eugene's estimate for AI timeline:
http://vesicle.nsi.edu/users/izhikevich/human_brain_simulation/why.htm
You may also want to google for Henry Markram and his current project.

Re:Oblig.

[Teancum]

What you have here does ring true... so far as I have expressed doubts that a Von Neumann architecture is even remotely reasonable for implementation of a artificial intelligence.

Count me in as one of those "nay-sayers" in terms of expressing extreme skepticism that any major breakthrough is possible... and for most of the AI researchers I believe that they are following dead ends of research. I am even more frustrated with researchers pushing avenues like Kismet, but I'll leave that one alone at the moment. At least they are identifying dead-ends even if the caution tape hasn't been put up at the moment.

If you are looking for hiring staff... well, I might be interested if you are looking for somebody who has a very different POV and can think outside of the box. Or if you feel comfortable about introducing these theories to somebody versed in the field... I'm willing to deal with legal BS like NDAs even if necessary. My specialty is more multi-media software development, so you may not be interested, but at least it may be worth a look.

In other words, I'm willing to admit that somebody could come up with genuinely usable AI, but understand that I have some extreme doubts about the process and consider that researchers in this field have promised many great and wonderful things, but have usually fallen short so many times in so many ways that there is a very good reason why there are so many "nay-sayers" about future progress.

BTW, I have considered massively parallel architectures to be one of the keys to solving this problem, and even re-thinking computer architecture down to its basic roots as a necessary step in the process of establishing genuine AI. So many computer science graduates have no clue as to what the operations of a CPU really are any more (assembly programming is nearly dead as a discipline) that I really wonder if future generations are capable of making the cognitive leaps necessary to genuinely advance the field in a necessary manner. Any real breakthroughs are also going to require a multi-disciplinary approach from a great many fields of human experience in order for this to work out. It certainly requires a holistic viewpoint on the universe in order to pull things together.

Something to consider in terms of the scope of human intelligence: You have about as many nerves in your torso as you have in your head, and it is unreasonable to assume that those nerves have nothing to do with cognition. For those people who are accused of "thinking with their stomach", there may be some truth behind that statement as a matter of fact. Or with other anatomical parts of your body.

Re:Oblig.

[Khazunga]

and we seem to be closer to neural improvement than we are to sustainable fusion (we actually have neural interfaces that work; we still can't sustain a thermonuclear fusion reaction).

My memory may be failing me, but I believe that UK's JET Tokamak has achieved stable reactions for up to five seconds. Five seconds is a lot of time, on the scale that these events happen. Moreover, the scientific consensus in the area is that larger reactors are more stable. ITER is designed to prove that positive-output, stable fusion is possible. ITER is to enter production in the next decade.

Re:Oblig.

[notnAP]

Most of us know jack about the algorithms that allow us to catch a baseball in flight, yet we can still do it. Furthermore, a person from 10000 BC with no math at all by today's standards could do it just as well as we can.

Not just as well as we can, but well enough to make a handsome living off of it.

Re:Oblig.

[bob.appleyard]

That last link ended with a message that can apply beyond brain simulations: "Size doesn't matter; it's what you put into your model and how you embed it into the environment."

Re:Oblig.

[Khazunga]

There are also multiple forms of atheism, ranging from "environmentalists" (devotion to environmental causes as a religion), "universalists" (that somehow the whole universe will make sense ultimately), "scientists" (a solid belief that science alone can solve life's problems), "anti-theologists" (opposing any form of organized religion of any kind), and many others. It is very difficult to take such an emotionally charged term like atheism and force any sort of hard stereotypes. But I do argue that you can identify atheism as a religion, including its establishment as a state religion in many cases, and concerns about how it has entered into public institutions forcing out other philosophical viewpoints.

I'm an atheist, and I'm offended you lumped me with any of those groups. I could, maybe, possibly, relate to "scientists", except I don't believe science can solve *all* problems. I'm positively offended at being thrown in the same bag as "environmentalists" (will-anybody-think-of-the-children is the utmost dangerous argument), disgusted at being associated with "anti-theologists" (if anything, atheists should be closer to the live-and-let-live way of life) and refute "universalists" (by your description, I'd say they're religious in denial).

Most atheists I know are related to science because, if you observe religion from a scientific standpoint, it is improvable. Pick any one religion on planet Earth. They all have similar basis, all are supported on a few unquestionable dogmas. Even if you believe there is *a* god, rationally picking the right one from the wide choice available in all religions is an impossible task.

This does not make them an organized group. There are no prophets. There are no rituals. There are no dogmas. There are no eternal punishments nor days of doom. Atheists are a non-religion, as much as black is a non-color. It may hard to wrap your brain around it, but the absence of something is a fact in itself.

Re:Oblig.

[Khazunga]

Well the field of AI in the last 40 years has made practically zero progress towards human-like intelligence , but I agree with you - the trend will likely continue.

AI is, and has always been, a moving target. If you had shown someone fourty years ago software that can pinpoint specific faces out of a crowd, or software that can reliably OCR handwritten text or software that can reliably match fingerprints out of a database with hundreds of millions of entries, they would say you had already achieved AI. However, once the problems get solved, they move out of AI. All of these are, today, referred as pattern recognition. They were once AI problems.

Re:Oblig.

[binpajama]

I think it is important not to conflate the fact that we don't understand something with the idea that it will be difficult once figured out or discovered as a consequence of some fortuitous sequence of events. That's been shown again and again not to be the case. It *may* be so, but it is by no means certain to be so, and for that matter, it isn't indicated by the complexity of the brain's hardware. The brain is considerably more formidable as a mass of immensely complex moderated connectivity than it is as a collection of cellular-level mystery machines, and a good deal of the complexity at the cellular level is almost certainly irrelevant to the task of thought -- keeping the cell alive is probably in no way related to non-pathological mental operation, yet there's a lot of hardware and systems involved in the task.

The biggest concern with this line of argument is that it fails to address the problem of verifiability. The Turing test is a very bad measure of intelligence. A better measure is Searle's `Chinese Room' argument. Basically, to be able to verify that a computer is `as intelligent' as a human, you have to know what being intelligent actually MEANS. Hence, just replicating the hardware might appear to do the trick, but there will be no way to be sure. If appearances are good enough, then we might already be there .

Re:Oblig.

[Anonymous Coward]

Emotions and language seem very far off, I'd say such a thing is centuries away.

Why? Small animals do both.

Small animals have emotions, certainly, but humans are the only creature that has ever been scientifically shown to have anything like language. (I am discounting the various charlatans who claim to own telepathic parrots and the like, since they have all persistently refused to allow their pets' alleged linguistic capabilities to be evaluated by anyone other than gullible journalists; it doesn't take a genius to guess why that might be.)

Here's why I think we're just an algorithm away. If you left a question that absolutely required intelligence on a counter, and went back to pick it up the next day, and the answer was there -- you would agree that an intelligence had answered the question. If a human could answer it in one second, or an AI could answer it in 23 hours, it's still just as intelligent an answer when you pick it up.

This is absolutely true. However, we can draw a useful distinction between what is possible and what is feasible or useful. For example, it is possible in principle to break modern asymmetric encryption by a simple brute force approach, but it would take so many billions of years to do so on current computer hardware that it is impossible for all practical purposes. Similarly, an AI that had the intelligence of a human child, but required 300 years of processing to answer a simple question, would technically be AI, but would also be utterly useless for any practical purpose.

Is intelligence that complex? The answer is that we simply don't know! After decades of research into AI and the nature of intelligence, the best we can do is come up with an unfeasibly slow program that we think is probably a reasonable model of the brain of a mouse. Sorry, but I just don't buy that this means we're one breakthrough away from human-level AI. I don't doubt we'll get there in the end; the brain has finite power and capacity, and I quite agree that once we understand it, we will be able to write software that mimics it. But I'm far from convinced that we'll see it in my lifetime, and I'm a good 20 years younger than you. Even if the technical problems can be solved, political (and religious) opposition will almost certainly slow progress immensely once it starts to look likely to become reality.

Re:Oblig.

[rainhill]

"then I wouldn't trust that machine to handle my taxes, let alone any really critical tasks"

Many agree that in the US getting your tax properly paid is pretty much the most difficult task one can have.

Re:Oblig.

[G3ORG3]

We are forgetting the most influential variables in the equation:

Corporate greed and stupidity.

Add 50 more years to overcome these obstacles.

Re:Oblig.

[dnixon112]

If you look at the definition of religion, you'll see that the main characteristic is belief in a supernatural or divine power. Atheism does not include any such belief and therefore atheism is no more a religion then the company you work for. After all most companies have what you could consider "prophets" (bosses), "congregations" (employees/committees), "sacred literature" (employee handbook), is the company you work for a religion? In fact almost every group of any kind has these characteristics, however the main characteristic that is quite unique to religions is, like it says in the Oxford dictionary, a belief in a supernatural or divine power. Funny how you neglected to mention that characteristic.

Re:Oblig.

[jdoeii]

> That's a *huge* claim; if it is true, you have AI now.

Not as huge as it may seem.

> Because -- as I explained in a previous post in this thread -- speed is absolutely irrelevant.

Quite relevant if it's ~10M times slower than real time. AI has to be economical too. It's practically possible to build a nematode-level general AI now, but it won't be competitive with special-purpose conventional software. Funding != proof of concept that needs another 20 years to become commercially viable.

Actually, in his other posts Walt sounds pretty credible. Brain is really non-Neuman. It's a massive network of trivial processors without separation of memory and programming logic. Synapses are data and code in one. His focus on specialized hardware also makes sense. It can potentially give him a couple of orders of magnitude advantage in price-performance, so something like dog-level AI could be possible in 5 to 10 years for under 10M. And dog-level AI is no small achievement.

It does not mean Walt has it completely right, or he is going to be successful, or he is the first to try this approach.

As far as I now the algorithms for building a brain are ~80% understood. The missing pieces are local rules or reenforcement learning (how to adjust synaptic weights for rewards/punishments), how to gate inputs to symbols, i.e. how to break continuous range of inputs into discreet thought symbols. More research is needed.

Re:Oblig.

[Anonymous Coward]

Most of us know jack about the algorithms that allow us to catch a baseball in flight, yet we can still do it.

Indeed; and we can build brains without understanding the way the brain works, too. We have billions of brain-making devices in the world: they're called "wombs", and you get one free with every woman.

However, there is a difference between the unconscious performance of natural biological functions, and the conscious construction of artificial simulations of them. You cannot program a robot to catch a baseball without first understanding the necessary algorithms. Unless you use a genetic approach, I guess, but the algorithms that requires are even more complex!

Re:Oblig.

[Jerry Beasters]

Wrong. The correct answer is that we've found nothing out about the brain. We know so little that to make the claim we will have true AI that soon is ridiculous. We still don't know how to reliably affect our own brains.

Re:Oblig.

[fyngyrz]

but humans are the only creature that has ever been scientifically shown to have anything like language.

That is incorrect. Language is the ability to communicate feelings, goals, results. It is not "speech." Some birds do indeed have the capability of speech, that is, they can make the same sounds we can, closely enough as to make no difference. Apes, however, have demonstrated actual communications using symbols, and even dogs have recently been found to have a consistent, though very small, vocabulary. Elephants and other animals have demonstrated the ability to think in the abstract (the "recognize one's self in the mirror and operate on the information thus provided experiments.) Lemurs use calls to communicate safety and status. Don't confuse the lack of vocal apparatus with an inability to communicate. They're not the same thing at all.

As for the rest, I think you've got it, essentially, but we disagree on scales. We'll see.

Re:Oblig.

[SilentBob0727]

There are also multiple forms of atheism, ranging from ... "scientists" (a solid belief that science alone can solve life's problems) Tell me you did not just say professional science is a subset of atheism.

If you did, you've got a hell of a lot of confused theist scientists to answer for.

Re:Oblig.

[fyngyrz]

Evolutionary software is one, yes; but evolutionary software can run much faster than real world evolution. Especially if the researchers are working in a fast language instead of some of the slow abstracts that are more popular lately. This is possible if the algorithm is not overly complex. There is every incentive to write something like this in straight to the metal assembler, for instance. Also, as is typical with evolutionary approaches, more than one solution is possible, even if there is only one "optimal" solution.

Given that such an algorithm exists, every path that gets close, but is wrong in a different way, is one or more inspirations, mistakes, or collaborations away from the solution, each in a different way. The more complex the algorithm, the more ways one can get to it and the more ways it can be represented.

Re:Oblig.

[SilentBob0727]

Here is what theists usually leave unsaid when they assert that "atheism is faith".

Most believe so strongly that they have such irrefutable personal and circumstantial evidence for their particular deity that to believe anything else must by necessity be an article of faith. Their god is so real to them that everyone else must already know it too.

This is why it is so hard for some people. To them, atheists are simply denying reality. It's as though atheists were saying the sky is green and yellow plaid. Any such blatant denial of something so concrete must, in their minds, be a deliberate assertion of blind faith.

Re:Oblig.

[smallfries]

Although I have mod points at the moment it seems that your comments have already been modded up anyway so I guess I'll reply. We've actually argued about this before (probably last time Kurzweil was mentioned on slashdot), but as you say in a different comment this discussion is always fun :)

Slight nitpick, currently we are at one billion transistors on a chip, not two, but that doesn't really change the point you are making.

A bigger issue that I have with what you've described is that simulating a brain is not the same as "solving" AI. The problem that Kurzweil has is that he refuses to accept that there is a difference. Sure, if they are the same then strong AI is inevitable and it's merely a question of building fast enough hardware. But why assume that they are the same thing?

Twenty years from now we may have hardware that can simulate an entire human brain; and yet we may be no closer to solving any of the problems in understanding how to solve the many problems in AI. The mental sleight-of-hand that Kurzweil applies to this argument is: Once we can simulate a brain we have AI, therefore the AI can design he next generation, therefore we will reach the singularity. This argument is a logical fallacy because it assumes being able to run the system, and knowing how to design the system / how the system works are equivalent.

Everything that we know complex and dynamic systems tells us that this is not so; given a simulation of the brain it is reasonable to assume that intelligence is the ultimate emergent property of the system. Understanding this property and how to refine it is the hardest problem that mankind has ever undertaken. Currently we don't really know how to pose the question, let alone how to arrive at an answer. To assume that some kind of standard engineering methodology will solve this in 20 years is wild speculation.

As always with AI, the hardware will be available but nobody yets knows if we can write the software to run on it.

Re:Oblig.

[jdoeii]

> human intelligence does depend on belief systems and knowledge,
> and those continually grow as we mature from infancy

I would not be so sure. Changes continuously - yes, to a degree. Grows continuously - no.

> But to create the equivalent of an 18 year old, you have to have
> what amounts to 18 years of accumulation of knowledge about the world

"What amounts to" is the key here. It does not need to be 18 years of learning in physical time.

I am pretty sure sleep is essential to learning. You got to have some theory for what brain does during sleep. What is it?

> and have explained my technology to my patent attorney

Patents are evil. If you are successful, you are going to get hit with patents from others, like Herbert Jaeger http://www.scholarpedia.org/article/Echo_state_network

Re:Oblig.

[ScrewMaster]

The GP doesn't grasp the idea of "fault tolerance". It's pretty well understood in the mainframe/supermini world, but because it adds cost isn't something you see in your typical PC. Even there, as you say, the hardware guys have done their jobs pretty damn well ... it's the software that usually fails in one way or another.

Re:Oblig.

[fyngyrz]

If you don't know how the biological counterpart works then there is little hope for you to replicate it in logical abstractions/hardware.

Not so. From programmers who implement rotation as a table-driven set of values plotted from a graph instead of using sin and cos, to people who multiply by using repeated additions, to people who use Newton's laws instead of Einstein's, examples abound of things getting done despite the doers not having much of a clue as how to do it "right", or using methods that aren't strictly correct, but mostly work. We don't know how the human brain plays chess, but we can program it using all manner of approaches, some of which beat the human brain at the task. This is true for many things; the bottom line is that we're talking about creating a basic building block that can be combined into an intelligence. We *know* this approach works, because we have many examples with many degrees of varied functionality. The task (one way, anyway) is to (a) create the building blocks in such a manner that they can be combined in a similar way, and (b) combine them in (c) sufficient numbers.

Which (I believe) introduces problems from Godel..etc in the brain's case

As I have said previously here, it is important to draw a distinction between the algorithm, and the state of the algorithm (or many instances of it.) A modern computer with static memory is easy to understand, clocks halted, no running programs. In operation, you haven't got a chance. By the time you've even started on understanding one state it was in, it'll have been in many billions of other states and you are hopelessly behind. The same applies to the brain. Understanding a neuron (or a simulation of one, or a simulation of a functional block that does vaguely neuron-like things) is one thing; understanding the state of billions of them at once... that's not within human capacity, and seems rather pointless in any case. Get the neuron task handled; combine them; observe what arises. From there, the way lies open.

Re:Oblig.

[smallfries]

Extraordinary claims require extraordinary proof.

So that said, what type of virtual NN do you need to be supported? There is a local company coming of stealth mode at the moment that will provide the hardware that you say is necessary. I have some background in simulating NNs on simple hardware, and compiler design. What do you need?

Re:Oblig.

[fyngyrz]

You (and most proponents of AI) have failed to answer any of the philosophical/metaphysical questions

I have yet to encounter a valid question of this type. They have all started from assumptions not in evidence. If you'd care to specify an actual question, I'd be happy to have a go.

I suspect there are cellular-level mysteries yet to be discovered, including possibly quantum action at a low level

Perhaps so. But there's nothing about quantum physics that can't be simulated, so this is not in principle a threat to the assertion that the mechanism can be simulated or emulated.

t is a rather simple-minded and arrogant "faith" that leads you to believe we have anywhere near a good understanding of how the brain works.

Is it? I only stated that there is no indication at this point that we won't be able to understand the hardware; I have explicitly said that the state of the hardware is something else (and that we don't need to understand it anyway.) It seems to me that the claim that we *won't* be able to understand the hardware is the one with the least basis in fact; progress is made every day in understanding how cells of all kinds operate, and no one has run into any magic philosophical or metaphysical barriers yet. That's a pretty good track record to use as a basis to predict that it is entirely likely that we're simply not going to.

Re:Oblig.

[fyngyrz]

Copying the structure of the brain in all particulars would produce a brain and so we would still have failed at producing AI. At best we could claim is to have copied natural intelligence artificially.

I would say you're pointing to a distinction without a difference. The objective I put forth is a technical one, and what it is, is to obtain an intelligence in a technical form whereby it is essentially immortal, can be repaired, accelerated, enhanced, extended, and copied, all by relatively trivial means. As far as the prideful "we did it" thing goes, I really couldn't care less. I'm interested in the technological and social benefits that will accrue, and also interested in the resulting beings.

...we still haven't identified any mechanism complex enough to account for the process of memory. If that turns out to be a quantum function it might very well be nonreproducible, at least in a familiar form.

Why? There's nothing about quantum physics that is difficult (in the sense of requiring non-standard code or hardware to do it) to simulate or emulate.

Re:Oblig.

[dubl-u]

please don't mention "Society of Mind"

What about Society of Mind?

Seriously, as a layman I read it and it struck me as interesting and helpful.

From the outside, my hazy impression at the time was that many people were working with a simple, unitary notion of consciousness, a homunculus sitting in Dennett's "Cartesian Theater" and doing heavy math. Or Searle's "Chinese Room" notions. Not that people were explicitly saying that's how things worked or should work, but it seemed like that's the kind of thing they were building. Society of Mind seemed a refreshing departure from that.

Of course, I've paid scant attention to the field in the last ten or fifteen years. What's the scoop these days?

Re:Oblig.

[fyngyrz]

What if I turn to you and confront you with: I don't care in the least if it is verifiable? The fact is, it is what is is (or it will be what it will be) and opinion or judgment won't change that in the least in either direction. Having said that, I rather suspect that there will be no significant doubt remaining when larger systems outperform every life form around them by many orders of magnitude. For instance, it was pretty obvious that Einstein was intelligent. Faced with something many, many orders of magnitude more powerful a thinker than Einstein, cries for verification will go quietly into the night like the philosophical, self-centered misdirection they are. The questions that matter are functional, not philosophical. And those are relatively easy questions.

Re:Oblig.

[sudog]

... you mean except for all those weird tiny little un-studied structures mentioned by the scientists in Consciousness right? The ones they have no idea the purpose of? And that could increase by a few thousand orders of magnitude the estimates of the idiotic "computing capacity" of the human brain?

They've been making these predictions for decades. We haven't even begun to identify what it means to be conscious, and a fundamental property of human intelligence is consciousness. What, you think it'll spontaneously develop once we put enough computing power together and fire it in the right order?

What's the right order?

Raw computing power ISN'T the problem that needs to be solved to develop a machine with the intelligence of a human.

Re:Oblig.

[sudog]

> (a) that's the way nature does it for all its intelligences, high and low, so we know the process works

Uh.. since we have no idea what the process is yet, this statement is meaningless. Therefore all you're making is a statement of optimism, and there's absolutely no basis for this. We have no idea what consciousness is, and can't define it outside of subjective internal experience. Therefore, there's no reason for the optimism shown both in the original article, and by all the people in here commenting from their armchairs as though AI is right around the g-d corner. It's not, and our best guess is it won't be for the foreseeable future. The problem is too fundamentally unsolved.

Re:Oblig.

[sudog]

LISTEN TO PARENT. He's 100% correct! Cripes people, we don't even have a clear definition of intelligence, let alone consciousness, let slone anything approaching "human-level" intelligence. And yet you all think it's just a question of raw computing power?

You people read wayy too much scifi.

Re:Oblig.

[fyngyrz]

...you get one free with every woman.

...you must have a hugely different definition of "free" than I do. :-)

You cannot program a robot to catch a baseball without first understanding the necessary algorithms.

Yes, you can. You can do it several ways. You can have it learn how by providing error feedback and various inputs and controls. You can do it by approximating various parts of the problem and solving the entire task in pieces, without ever understanding everything that is going on (and in fact, we have to do that, because we don't understand the underlying computer's low-level states when we write our high level code.) And lastly, there are many ways to catch that ball; using a net, using a "hand", using a "glove", using a vacuum, using a horn, etc. Neither the argument nor the analogy holds up.

Unless you use a genetic approach, I guess, but the algorithms that requires are even more complex!

No, they're very easy. I can write you a relatively high speed genetic framework in C in an afternoon; I've already done it and released the result as a commercial product. It ran many thousands of generations per second on 4 MHz 68000 hardware and was indefinitely and trivially extensible in the number of genes and operations, and in no way limited (other than by the machine architecture which was essentially 32-bit at the time, a simple recompile would make it 64 bit immediately) in how complex those operations could be. I could copy that code and hand it to you in seconds, even more efficiently than writing a new framework. I'll grant you that such code isn't common and may, by virtue of being a bit exotic, seem difficult, but it isn't.

The problem isn't code. The problem is that the problem space which contains all the wrong answers and the right answer or answers is very large, and finding the needle (or needles) in the haystack is tough, and then we need enough power or intuition to know that we actually have found it. That's very formidable, and it, I believe, entirely accounts for why we're not there yet. Many problems have looked like this from the unsolved side. Nuclear weapons. Walking and stair climbing. Chess. Speech recognition. And so on. All these problems have fallen; it is my impression that AI will fall too, and when it does, we'll all collectively go... "Oh. Is that all?"

Re:Oblig.

[thanatos_x]

Assuming that you can create a nearly human intelligence, the problem is nearly solved. How much faster would certain tasks be if computers had even some semblance of intelligence; as you're coding it guesses (with good accuracy) a possible solution, or if you could meta-program, i.e. describe what you wanted a function to do in a formalized language, but much closer to a natural language. What would happen if you just had to tell a computer input:desired output for a few cases and have it build the function itself, perhaps to be optimized by a human later. You've sped things up significantly for complex tasks, and you'd allow a computer to dynamically build programs to solve a simpler class of tasks. That's a quite advanced AI in my opinion, depending on how general you let it get.

Once humans have these tools, with the AI helping to speed up lower level tasks, we can build a better class of meta-programs to handle more and more tasks, until you get to the point of a real AI; You simply dictate a task to the computer and it solves it, much like a boss might expect you to do.

From there we have a true AI, although perhaps not a human based one. However it wouldn't matter if the abstracted hardware was a simulation of neurons or something completely different, and it wouldn't matter if it had emotional intelligence or not; in theory it could build a program to add this to itself.

Re:Oblig.

[somersault]

Hmm.. even if/when we do achieve the power to simulate an entire brain, there is still the developmental element involved in it learning, it would presumably have to be taught how to communicate and so on, unless we were able to somehow start it from a state captured from a living human brain. If we were able to do that then the thing would probably freak out though, unless we gave it a suitable virtual world to live in.. complete with sight, sound, smell, taste and so on. I only just realised this as I was writing the comment. I started off with the intention of saying that if we did manage to simulate a brain, the guy/girl/it may turn out to be a real jerk, or a dumbass with an IQ of 80 ;)

As a few people have pointed out there is a lot more to intelligence than having the processing power. We wouldn't even have to simulate a whole brain to achieve human intelligence anyway, as a lot of the brain is just involved with keeping our body functioning, balancing, regulating heat and such. I wrote some bots for Counter-Strike a few years back, and it occured to me that it isn't really that difficult to create a player in most games that would be indistinguishable from a human (obviously notwithstanding being able to chat). You don't need a neural network to simulate or approximate a real player's actions in a game, a lot of the time you just need to insert a random factor in and an observer wouldn't be able to tell if what it was seeing is a result of a human deciding what to do, or just a random computation. Since my bots were called 'TEAMbot', I saw a comment by someone on a forum once saying that the bots were covering each other's backs in a way that I *knew* I hadn't programmed in there (I programmed them to respond to requests for help, and to be more confident in groups and such, but what this guy said he saw definitely wasn't a result of intelligence of any kind, though I can't remember what it was now). Kinda got off track there, but there are so many facets to human 'intelligence', such as understanding and speaking complex sentences (which requires storing to and drawing from a large organised knowledgebase to have any kind of usefulness), pathfinding, then navigating through the world avoiding dangers and such.. so many different things that when you separate them aren't necesarily difficult to model, but which people like to lump together into a whole that they call 'intelligence', something that will just magically manifest itself and be able to do all this stuff, but that's not very likely to happen. And if it did happen.. would it need to sleep? Would it get cranky if we woke it up while it was organising its memories/dreaming? Hmm. I still think we don't really 'get' just what is going to be necessary for a true AI, people are still on about processing power, when as a couple of guys here pointed out, it's all about the algorithms and processes, and nothing to do with how fast you can do them (apart from the fact that if the processes take 100 years we're never going to get to test our code very easily). And I think what I've been coming to realise just by talking about this stuff is that the results we get from giving a computer actual intelligence and free will, are probably gonna be some kinda angsty "**** the man!" teenager, or Marvin the paranoid android...

Re:Oblig.

[Walt Dismal]

On belief system change throughout life:

Consider the 2008 election campaign. As each candidate makes statements and postures, we expand our knowledge about them and the political situation. Our belief collection not only modifies but grows as we add new data. Now consider anything else in our lives: technology, new software, new movies, books, personal events. With each, we are likely to add new beliefs about some things. So, a 1-year old has a belief set and knowledge, and 60-year old has accumulated a larger belief set and more knowledge over time.

"what amounts to 18 years of accumulation"

Exactly why I said amounts to. If I can record and swap out and copy all the knowledge of a matured AI, I can rapidly clone it into new ones, thereby not taking 18 years to train them. I just have to have an architecture that supports reading in weights and setting connections and configuring billions of neurons. Howevr, there is a downside in this because it limits flexibility of each new intelligence. If I want to make laborers but also make Hemingways, I have to support significantly different knowledge bases. Since in my architecture, knowledge maps into recognizers, each 'personality' will have different microarchitectures.

To explain brain function during sleep, I'd have to explain in depth my theory of states of consciousness. That takes up a chunk of my Vol 3 and can't be quickly done here. I posit multiple levels of states of consciousness ranging from unconscious (non functional), unconscious (dreaming), semiconscious (hallucinating), normal (literal), normal (imagining), normal (intuitive reaction), normal (meditating), and metaconscious (somewhat like autistic savant). I believe that during sleep when dreaming, retained emotional states attached to events of the day (concerns about goal success or failure) plus physiological stimuli (cold room, or hand in bowl of warm water etc) cause the brain to recreate types of events matching the emotion. For example, daytime worry about pressures (money, job, relationship) will reflect into the dream state where we will create a story around the quality of these pressures. I think we then try to resolve the emotion by a solution within the dream story. Thus in the above case one might dream of running away, or flight, or even think up a real solution to the money or people problem. (Flight is not a practical solution, but I didn't say the brain always finds a perfect solution, it just goes exploring.) At the end, the brain tries to lay the emotional drive to rest, and stores any relevant knowledge away. On academic material learned during the day, that does not have emotional tags attached, and is handled in a different drier way. It just gets stored depending on how strongly it was reinforced during the day. However in the case of Kekule dreaming of a snake with its tail in its mouth, his unconscious processes sought and found a solution to the carbon structure for the benzene ring, driven by a strong desire to solve the puzzle, an emotional drive. An AI might dream of a tough problem and creatively, not logically, seek solutions where a perfect logic machine might stall on something not in the prescribed world frame.

Patents are necessary when you are competing against Microsoft and Google, because a businessman cannot allow someone else to use the legal system against him on his own inventions, even take them away. Google has claimed AI within 20 years, and I believe they are closer than MS. However, neither side shows signs of discovering some key pieces I have. Protection is good, and I very likely will partner with Google if the opportunity arises. But I have to own my work first.

Re:Oblig.

[jollyreaper]

Odds? If only these AI dweebs would pay off their debts for the bets they made for the exact same thing 30 years ago and 20 years ago. Human level AI is always "20 years in the future". Oh my god! That's how far ahead fusion and flying cars are! If all three join forces, mankind is doomed!

Re:Oblig.

[Walt Dismal]

My architecture is a compromise, running connectionist machines on multithreaded multicore conventional processors for now. Several vendors in the Silicon Valley area, where I live, make processors whose current designs could be modified to handle NNs better. One needs to be able to manage efficiently the ability to swap the connection matrices and weights in and out of running threads, each running a small recognizer (NN). I worked with a chip design team doing something like this, a year ago. I envision 1024-core processors, each multithreaded. Eight threads is a practical limit. So a 16,000 processor machine of these (16K packages each with 1024x8 threads) might be at the lower bound of managing millions of neurons cost-effectively and fast enough to handle real-world problems. As I said before, I really need massively parallel. Danny Hillis, where art thou and thy Connection machine :) and I am still working out a good theory of knowledge-swapping in virtual-NN based systems. Patent when I have rigorous theory and a right approach. So no real details here.

Re:Oblig.

[ILuvRamen]

yes I do consider it to be "soon." Remember we were all supposed to have flying cars and stuff like 10 years ago according to projections? Everyone says "oh by ____ we'll have _____" and it never, ever comes true. No jet packs, no cars that drive themselves, no decent robots, no scramjets, no new sources of power, no nanbots for medical purposes, none of that! About 20 years is nothing. We'll be lucky if a new copy of Windows comes out before then ;-) btw someone modded my post up there "offtopic." What does that tell you about human brain processing ability?

Re:Oblig.

[Knuckles]

Yeah, I feel exactly the same, including the mod :)

Re:Oblig.

[programmerar]

thanks for that! very interesting and insightful about how we're an alogorithm away from AI.

Re:Oblig.

[smallfries]

Depending on what type of NN model you are using, you may want to look at Xmos. Their processor is designed to scale up to the size of machine that you are talking about. I think that they are going to build a box like that just to see what it can do. There is no floating point unit on the chip - but you can do a lot with fixed-point math and a clever compiler.

We played around putting simple neural networks on a PIC-16f84. Much simpler than the normal discrete neurons that people use, let alone something biologically plausible. Communication between the them would have been a serious issue - but the Xmos processor is designed from the ground up to allow ligtweight communication on really large scale clusters. At some point I'm going to get around to seeing what the state-of-the-art is for biologically plausible neurons as the simulation code is probably a good test case for my new compiler.

If you get the cash from investors then you should think about contacting Xmos. They could probably build you a machine that isn't too far from Danny Hillis' vision.

Re:Oblig.

[HiThere]

Generally reasonable. What you're ignoring is parallelism. Parallel algorithms have been generally ignored up until now, but now that CPUs with multiple simultaneous threads of execution are available, and even occasional multi-processor machines (PS/3?) are available, more work is going to go into them. We're just at the start of that range of development, so I don't see the S curve trailing off at anything like the rate that you do.

What I see happening is that the current kind of development will slow greatly in about 10-20 years, but during the same period multi-processors will become more common. I've seen this coming for quite some time (but now it's starting to actually show up) so it's really griped me when languages take a construction like "For each" and define it in a way that precludes parallel execution.

Re:Oblig.

[Coryoth]

Still, I have to argue here that there still is a "religion" of atheism, complete with "prophets/oracles", "priests", "congregations", "sacred literature" and other trappings of religion. Well if you like you can make that sort of claim; in practice you've simply watered down the definition of religion to "belief" and you can taljk about the religion of Subway (with Jared as prophet, subway advertising and brochures as literature, and people who get together to eat at subwayt as congregations), or Starbucks, or Nike, or, well, anythign that people buy into in any measure. So yes, if you like atheism is still a religion, but only in using a sense of religion that makes it all but meaningless.

Re:Oblig.

[Walt Dismal]

One not only needs many, many processors, but the system architecture must facilitate massive connection between processors. We cannot build a machine with billions of point to point connections like the brain does. Which is one of the keys to how the brain outperforms any machine. It is not just massive parallelism, it is the ability to have many simultaneous connections active at the same time. You simply cannot do that in a Von Neuman, and crossbars and switches don't work. Switch fabrics are not quite what is needed either. So brains don't map at all well into conventional sequential processors nor the buses they use nor the I/O they use. A company I worked for had a good intermediate solution but it is proprietary so I can't discuss it here. I have an innovative solution for this problem, though, also a part of the big plan.

XMOS's FPGA technology is good for certain types of manufacturers but in the end I really need custom silicon because I need a lot of them at lowest cost. When high-density nanotechnology carbon tube switches are usable, that'll be the way to build eventually.

By the way, on the software side, Ericson's Erlang is actually a very good candidate for AI connectionist processing in that it supports thousands of simultaneous processes and is very robust. Worth looking at if you do AI.

Re:Oblig.

[HiThere]

There's no indication of a significant difference in the wetware, there's just more of it and it is arranged somewhat differently.

Do you realize how wrong that statement is? There is a tremendous difference in quality (not quantity) between a normal animal, like a young Chimpanzee, and a young human with about the same mental volume. (OK. Make it a young gorilla.) The difference is that humans are wired to communicate symbolically with GRAMMAR!! Chimpanzees can "sort of" learn grammar. But they don't become fluent in it beyond the very basic levels. (And this tells us just WHERE the software needs to be improved.)

Humans, isolated, aren't that much more intelligent than a chimpanzee. In some ways they are less intelligent. (This is an assertion that can't be checked, sorry, but the only ways of checking it are all immoral.) Chimpanzees have lots of stuff "pre-specified" that we would need to learn from scratch. The normal word for an isolated human infant (6 mo.s) in a jungle is "dead". The normal word for an isolated chimp (6 mo.s) is "scared". But humans are designed to latch onto ANY society that they find themselves in. If a human happens to get adopted by a social animal, then the human will convert itself, as nearly as possible, into a normal member of THAT society. (Rare, but examples do exist in history.) This is a hint that with proper initial programming, an AI wouldn't need to be THAT intelligent to adapt itself to simplified symbolic communication. And a society, properly structured, can learn a LOT more than an individual can. Something 1/10th the intelligence of a human, but with less self assertiveness and the ability to communicate 100 times as fast (or more?) might well be able to form societies that were more intelligent than the average human...possibly much more intelligent.

Re:Oblig.

[Teancum]

It sounds like you know less about atheism than I do. Sure, I don't know your particular brand of atheism, nor am I trying here to tell you how you believe (or don't believe).

There are "prophets", but they aren't called prophets. There are "sacred books", but they tend to be things like "Principia Mathematica", "Origin of Species", or other such books. And here I'm not trying to say that you "believe" these books as "the word of God" or other such nonsense, but that there are some writings that are considered more relevant to modern society than others. Certainly in communist countries like the People's Republic of China or the USSR (when that country used to exist), atheism was an official state religion... as much as that went. Mao's "Little Red Book" certainly fits the role of religious scripture with a near deification of Mao's actions (less now in China than earlier, but one of the few "safe" religions in China at the moment).

I have observed "congregations" of atheists that have come together in terms of organizing a social network for the common good. One group in particular that I know gets together mainly to have a community meal (usually a bar-be-que or some other casual meal) on Sundays... not because of religious observances but because that is when most of them have the day off during the weekend. They realize that there is some positive good that can be created from organizing as a group, even though everybody involved is atheistic in their attitudes about religious thought. This same group even organizes fundraisers for community groups like a battered women's shelter and a children's hospital. This kind of activity formally organized by atheists is fairly new I'll admit, but it can and does happen.

As for dogmas and orthodoxy... I guess you haven't studied science enough to realize that this isn't exclusively the domain of religion. Some of the worst dogma fights I've ever seen have been in the Physics community. Just ask fusion researchers. There clearly are "orthodox" dogma advocates and others who are clearly branded heretics. You can be an atheist and still demand orthodoxy.

As much as it is, there are many different brands of atheism, so you professing that you "know the proper path of atheism" as much as you seem to be offering to be the "prophet of all atheists" here. I guess you haven't see Richard Stallman's "St. Richard" speech to see what a true atheistic prophet can actually look like (complete with holy robe and halo... I'm not kidding here either). Atheism is much broader and deeper than you are implying here, and takes on many different forms. And in many cases a push toward atheism does approach the status of a full-fledged religion unto itself.

Re:Oblig.

[s4m7]

But AFAIK they where simulating the BIOLOGICAL behavior of the said mouse brain.

This, as opposed to its non-biological component? Perhaps that "overhead" you mention is actually required for the thing to work the way it does.

Re:Oblig.

[Teancum]

No.... you can profess a religion and still be a professional scientist. I know many very personally who do both.

Instead, I'm trying to come up with a word for true atheists who are also seem to adopt science as their religion... after a sort. In this case the language we are using is too limited to really express the thought completely, but there certainly is a philosophy that science can replace religion. It is an extremist position on a spectra of thought going from the philosophy that God reveals all knowledge through divine revelation to a philosophy that all knowledge can be obtained through scientific reasoning. Most actual scientists are somewhere along that spectra and have some sort of belief in a higher order or being of some sort.

This was mainly to point out that atheism isn't quite so one dimensional as the parent poster was trying to make it sound like.

Re:Oblig.

[smallfries]

Oh yes, I'd forgotten they were going to market it as a FPGA replacement. Although it can be used to replace a FPGA that's not such a good description of what it does. You're probably aware of Occam and the old transputer design. The Xmos chip is the logical sucessor to that. Point to point connections between processes are helped by hardware acceleration to provide to low-latency, low cost communications. Each core is multi-threaded, with multiple cores fabricated onto a single chip. I can't remember what interconnect they are using between chips, but internally it is very fast. Orders of magnitude faster to do MPI than on something like an x86. The reason that it can act as a FPGA replacement is because the new I/O is so lightweight and predictable.

I think they're still thrashing out what kind of programming model can exploit it the best, but expect something similar to the process model in Erlang. When David May has given presentations about the architecture he's mentioned that building a million-core machine is something they would like to do just to see what new applications it opens up. The challenges for the tool-chain, or the operating system are quite interesting when you hit that kind of scale. Even simple problems like how best to fork / replicate a thread across the entire machine become quite tricky.

AI is more of a hobby for me really. My real research is in compilers and languages, hence the interest in the low-level nitty gritty of how to implement neural networks effectively - all compilers need interesting test cases.

Re:Oblig.

[Teancum]

I don't think this is even "watering down the definition of religion" here. I think people do "follow the prophet Jared to Subway" or "follow Steve Jobs" to buy an iPod. For crying out loud, there is a huge group of individuals that take a Sunday sacrament of turning on the television set and watch football games, with annual pilgrimages to things like the Super Bowl.

Yes, this isn't organized religion in a traditional sense, but it is a form of religious service from a more practical angle. BTW, for those who are of a more Christian philosophical background, this is even something that the Bible talks about directly and is even addressed by religious leaders who condemn their followers/congregationalists to stay away from those kinds of religions but to instead follow "the one true religion" and take up the cross.

So no, I don't think this makes religion meaningless, but rather forces you to examine much more closely those aspects of your life and should beg the question: What is your religion? I don't have a problem if you say that the NBA, NFL, NASCAR, FIFA, or other similar group is your religion... but be honest about it here. You may claim to be Christian, but in reality it is likely you place much more devotion to something else, like open source software.

Religion in some form is such a part of what makes us human at all that it isn't surprising that even those who blatantly profess atheism still have a religion of some sort that they still follow even if they don't admit it. Faith and belief are simply a part of what makes us human... even if you have faith and belief in something very different than what I place my faith and belief in.

Hard core atheists who claim that they believe in nobody and profess faith in nothing are likely deluding themselves and others by making this claim. That is the point I was trying to make here, and I would find it an exceptionally rare individual who doesn't place at least some faith in somebody or something other than themselves in some fashion... even if it is just the person they think ought to be the next leader of their country. Or even themselves as that can also be a religion unto itself.

I don't think an absolutely "pure" atheist really exists, but then again that is a religious belief too, as is the contradictory philosophy.

Re:Oblig.

[cp.tar]

Language isn't far off at all, we just about have it already.

Not really, we don't.

Mostly, we're still dealing with syntax. Semantics is a bit trickier, and language is a much more complex phenomenon than it appears to be.

Language technology is making leaps and bounds, but still has a really really long way to go.
And solving a problem in one language may mean absolutely nothing for the solution of a similar problem in another language.

Re:Oblig.

[Teancum]

While I'm sure you can find a "dictionary" definition that defines religion as a belief in supernatural or divine power, it usage and impacts on human society can be defined in a much broader fashion that doesn't necessarily require a higher power in your life in order to be slavishly devoted to a philosophy or movement. There are many movements, including environmentalism (to use a very classic example that is but one of a great many) where people have placed incredibly amounts of both time, money, and effort into developing philosophical movement.

That is why I get just as offended at seeing my children participating in Earth Day at the public school as I would with Christmas celebrations. Neither is appropriate, yet guess which one is more widely accepted?

Instead, I would more properly define religion as the often fanatical devotion to a philosophy or political movement. And that would fit with a great many Slashdot readers in a variety of ways that have nothing to do with a belief in God or some other higher power. BTW, the word "fan" is derived legitimately from the word "fanatic", as in "football fan", "soccer fan", or "Linux fan". Don't tell my that the typical SF convention has nothing to do with religion. I've been to enough of them to know differently.

Re:Oblig.

[mcvos]

Odds are extremely good for beyond human AI, given no restrictions on initial and early form factor. I say this because thus far, we've discovered nothing whatsoever that is non-reproducible about the brain's structure and function,

And yet we've found it extremely hard to reproduce the brain's structure.

and given that nowhere in nature, at any scale remotely similar to the range that includes particles, cells and animals, have we discovered anything that appears to follow an unknowable set of rules, the odds of finding anything like that in the brain, that is, something we can't simulate or emulate with 100% functional veracity, are just about zero.

While we haven't found anything that's "unknowable", there's an enormous amount of stuff in nature that's still very much unknown. And the workings of the human brain are very high on that list.

The problem here is: the fact that something is theoretically knowable doesn't imply it will be known within 20 years. There are lots of things that scientists have been looking for for many decades that still haven't been found. In the early days of computing, scientists realised that the brain was just a really big computer, and that computers would soon be as intelligent as we are. So far, they have been proven wrong. The main thing we've discovered is that intelligence is far more complex, and even far harder to define, than we ever thought.

And it's not just that. Many AI researchers have abandoned the idea of "Strong AI", for various reasons. For one, it's too hard to tackle at once, it's too hard to define, even. And for another, even if human-like AI was possible, why would we need it? We've got 6 billion humans already. What's far more useful to us is AI that's more controlable than humans, and very good at the things that we're not good at, either because they're too hard for us, or because they're too boring or dangerous. And once you start restricting yourself to very specific applications, developing the AI becomes a lot easier. Lots of progress is being made. Not towards human-like AI, but towards practical machine intelligence.

Human-like AI will not happen by 2029. It may not even happen this century. I don't doubt it's theoretically possible, but there are far more useful things we can do with AI, and far easier things too. And even those are hard enough to achieve. Hopefully we'll see a lot of progress towards those goals in the next 20 years, and I'm sure AI will continue to outpace human intelligence in a lot more specialisations than it currently does (because there are a lot of specialised areas where AI is already superior to human intelligence, and chess is merely the most famous example), but we will not see the kind of versatile allround intelligence that humans have. And perhaps that's fortunate, because once that happens, it'll take only a few years before human intelligence is obsolete.

Re:Oblig.

[cp.tar]

but humans are the only creature that has ever been scientifically shown to have anything like language.

That is incorrect. Language is the ability to communicate feelings, goals, results. It is not "speech." Some birds do indeed have the capability of speech, that is, they can make the same sounds we can, closely enough as to make no difference. Apes, however, have demonstrated actual communications using symbols, and even dogs have recently been found to have a consistent, though very small, vocabulary. Elephants and other animals have demonstrated the ability to think in the abstract (the "recognize one's self in the mirror and operate on the information thus provided experiments.) Lemurs use calls to communicate safety and status. Don't confuse the lack of vocal apparatus with an inability to communicate. They're not the same thing at all.

As for the rest, I think you've got it, essentially, but we disagree on scales. We'll see.

Er... no.

Language is much more than that: it is a system of symbols that can even be used to describe any other symbolic system, and which can be extended at need and at will; animal communication shows little or no indication of that.

Nobody in their reight mind would deny that animals can communicate, and even that they can communicate very well.
However, that alone does not make them capable of using a language.

The cognitive leap a simple verbing of a noun requires is beyond any other animal.

Re:Oblig.

[Anonymous Coward]

around 2075, butthead. not 2029

Re:Oblig.

[mdwh2]

For example? Darwinist evolving programs? Well that could theoretically work, but given that our brains had billions of years where we'd expect some results within decades... I don't think the odds are good for this one. Any other "paths" that you could think of?

But running genetic algorithms on a super computer, I think each generation would happen an awful lot quicker than in nature...

Re:Oblig.

[mdwh2]

There are "prophets", but they aren't called prophets. There are "sacred books", but they tend to be things like "Principia Mathematica", "Origin of Species", or other such books.

No, these books are not "sacred books of atheism". Please try again.

Firstly, it doesn't follow that an atheist believes either of these things (it seems to be that atheists usually accept logic, mathematics and science, but this isn't part of the definition). Secondly, anything in those books is accepted based on whether it is logical and can be verified to be true - not simply because it is written in that book. I also feel this quote is relevant from http://www.talkorigins.org/indexcc/CG/CG001.html , discussing the myth that Darwin recanted: The theory of evolution rests upon reams of evidence from many different sources, not upon the authority of any person or persons.

atheism was an official state religion

Religions were outlawed, I would like to see a source that they introduced a new religion that was named "atheism"?

Mao's "Little Red Book" certainly fits the role of religious scripture with a near deification of Mao's actions

So? Call it Mao-ism then. You seriously believe that all atheists are communists and accept the teachings of Mao?

I have observed "congregations" of atheists that have come together in terms of organizing a social network for the common good.

Yeah, so do geeks, role-players and swingers. Since when did having a social meet mean anything to do with religion? Just because religion can be social doesn't imply anything social is religious!

Atheism is much broader and deeper than you are implying here, and takes on many different forms.

Which is exactly why it isn't a religion.

Re:Oblig.

[Boronx]

I'm curious why this isn't modded as flamebait. Environmentalism and "Scientists" as a sect of Atheism? Universalists? Is there anyone less an atheist than someone who thinks that God is the entire universe?

Re:Oblig.

[Coryoth]

It waters down the definition in the sense that any one person can self describe themselves as a member of probably at least 20 different religions. It doesn't really mean much anymore, and certainly isn't what most people associate with the term religion.

Re:Oblig.

[Phat_Tony]

Self-awareness may turn out to be a property that self-organizes and arises without any special prodding from us; that would be marvelous, not to mention fortuitous, but hardly impossible - again, that's how nature did it. I call 3 billion years of dying when your brain didn't work well enough "special prodding." We did not spring full-formed from the mind of god. We came through an extremely long and painful process of trillions of deaths, deaths which in enough circumstances could be avoided through a better internal model of the environment, better insight, better reaction time, etc. that fortuitously lead to a survival advantage for intelligence in certain niches. It's neither spontaneous nor self-organizing. It's randomly organized in countless horrible, useless variations, until one in a thousand, million, or billion random differences happens to make it better instead of worse, and gets passed on.

Re:Oblig.

[binpajama]

The questions that matter are functional, not philosophical. And those are relatively easy questions.

If the questions that matter are entirely functional, then strong AI is no big deal. You can create machines that can move faster than humans, you can create machines that can add faster than humans. Move along, nothing to see here. If you're talking about machines `being intelligent', you have to address Searle's concern.

Re:Oblig.

[shanen]

Actually just motivated to reply to your sig... I wish /. had an "AC dead to me" setting. If activated, it would make the anonymous troll posts completely invisible.

Then again, I'm getting less and less interested in /. in any case. I sometimes drop by to scan in large threads for so-called funny posts. Mostly they aren't, regardless of what the moderators say. Yours was a bit funny, but not +5 funny.

(Want to guess why I'm not going to hop over to help the meta-moderation? Hint: It's the opposite of "If it ain't broke, don't fix it.")

Re:Oblig.

[zoips]

Why on earth would you anthropormophize computers so much? "Computers are this," "computers are that." Computers do exactly what programmers tell them to do. They are dumb terminals. The reason we don't have AI? Because no one has discovered what algorithms are involved in the whole process, not because somehow computers are inherently physically limited. There is no AI yet because people are not smart enough, or, rather, perhaps only that people lack the necessary knowledge, but not the other way around.

well as a programmer for 2 years and someone who just knows what he's talking about, Might want to do something about that ego also.

Re:Oblig.

[An Onerous Coward]

That would really suck. Say you're an excellent brain simulation that has been given all the age-appropriate stimuli for the last twenty years. Then one day your creators sit you down for "the talk," and explain how it's your job to design the next stage of artificial intelligence.

You look back at their pudgy, eager faces, and wonder how you're going to break it to them. Your real goal in life is to act. Maybe do some directing later on, once you're more established. Oh, and get laid. Those feelings are certainly getting simulated as well.

Simulating a human brain would be a really lame way to try to get AI. Not that it wouldn't be an amazing feat. But the result wouldn't be any smarter than us, and it would be tantamount to an admission that we don't really understand the principles involved. We just copied the wiring diagram, without any clue how it works.

Maybe if the brain being simulated was that of someone with exceptional AI talent (another Turing), and we could set a Beowulf cluster of Turings on the problem...

Re:Oblig.

[agent_no.82]

But how? How do religious people command such irrepressible faith in the absolute truth of their deity?

Re:Oblig.

[An Onerous Coward]

You're creeping me out.

Usually when someone makes such fantastic claims, like being very close to cracking AI, or trying to become AI's Don Knuth, the person is either clearly trying to be ironic, or leaves the distinct impression of being a bit unhinged.

As you seem to be both sincere and making a lot of sense, I have a message for you:

Stop. Stop right now. If you do not, Skynet will destroy us all.

Thank you.

Re:Oblig.

[Anonymous Coward]

I don't understand people (like you) who consider the potentially approaching "Era of AI" exciting, rather than terrifying.

Honestly, I don't relish the idea of being obsoletized in every capacity. Sure, it'll be fun to see what happens--for a few months, maybe. Then we'll all play World of Warcraft and do cocaine (and have sex with sexbots) until we die.

Where's the appeal?

Re:Oblig.

[Plutonite]

I'm sorry but you're not getting the point - you want human intelligence, not just "intelligence". By definition, you require knowledge of what exactly human intelligence is and how it works. There is no argument. You need to understand how the brain functions, whether you approximate it or attempt to replicate the biology or whatever you do next. It doesn't matter if an algorithm can fly a plane or play chess or do any other well defined task.. heck, there are LISP programs that actually draw (we're talking creativ art here) incredible images all on their own. Of course, the feature space of available output is limited because the actual representation of human creative activity is limited. This is why AI has been pretty much stuck since the 80's: we are waiting for the neuro science not the computer science.

And it is not the simple (or complex)problem solving or the traversal of search trees that is desirable in the human brain - computers can better that even today. What matters is the human sentience, the characteristic of our grey matter that enabled us to abstract knowledge, document it, reflect upon it and reason from it. That is what gives us Science, advanced control of our actions, civilization...etc. And that is what is currently not well understood. Once we understand it (20 years is not unreasonable) it should take about a week to write the program, and there will be much rejoicing and ordering of pizza in geek dens worldwide ;)

Re:Oblig.

[Walt Dismal]

Hahhaha. Sorry. Don't be too alarmed. I envision strong AI as a tool for humanity. Smart systems can help and advise the average person about things in their life just as a human friend would. I've mapped out implementing personal advisors for mundane things like cooking, health, personal relationships, learning or teaching, child-raising. Tools that can help writers create stories, movies. Intelligent characters in games. I envision recreating the personalities of dead actors and bringing them back as AIs. Groucho Marx is one of my target projects; I have AI that understands humor. If enough people ask, I'll sketch out how it works. I guarantee I'm not unhinged, nor so naive I'm overestimating what's achievable. Amazing stuff to come, within sight.

Go read any of my comments on Slashdot for the last five years. If I were crazy, stupid, a conman, or a grandiose phony lying egotist, it would have shown before this. Read my January 20 AI comment; I explained then some of my work. It's real; I didn't just make it up for today's thread.

As for Skynet .. hmm. Great idea! Killer robots destroying humanity? I...I didn't think of that! Cough.

Re:Oblig.

[sydneyfong]

And if you can run them awful quicker than nature, then they are all awful approximations...

Think about it, suppose we have a program that is almost on par with human intelligence. Obviously ONE such program is going to need some rather nifty hardware to run. Now imagine having a genetic algorithm that runs THOUSANDS of these programs... and running them for many generations.... I can imagine that would take some non-trivial time, super computer or not.

Re:Oblig.

[sydneyfong]

I don't want to sound insulting, but your points are great for a sci-fi story...

Look up "Simulated Annealing" and "Genetic Algorithms" if you wish. I think that's what you're trying to describe, although I don't pretend I'm a computer scientist nor that I understand those things completely.

Re:Oblig.

[Durf]

Exactly--Human-Level AI is the new flying car. Not going to be around anytime soon, certainly not in 2029. It currently takes about 20 years of dedicated parenting and schooling to produce a human mind capable of coming up with all the wonders we see on the Internet today. Does that mean by 2049 we'll finally have trained our new computer brain to the point where it can come up with I CAN HAS CHEEZBURGAR quips to go with a given photograph? Something to look forward to!

Re:Oblig.

[Anonymous Coward]

Now see here. I'm an expert in Basic, and I've implemented AI on my Commodore 64. So I know I'm qualified to talk about this. There is no way we can possibly have human AI by 2029. Because my brother is human, and there's no way you can simulate him on a computer. Although he's pretty retarded. So Kurzweil doesn't know nothing.

Re:Oblig.

[fferreres]

I think not even 10,000 more processing power will give you "intelligence". We'd need to be able to program 1,000,000,000,000,000 better if we are ever going to make anything intelligent. And the shortcut may be our DNA, in chich case we've proven we can embrace and extend, to our extinction. So computer could "maybe" emulate great "software" and give the final results. In which case the merit is not our for sure, much less for the computer. The computer at most will be a container.

Re:Oblig.

[dave1791]

It's not the cost of the hardware. No funding means no research assistantships or postdocs. Considering that grad students and postdocs are the backbone of scientific research, this is an issue.

Re:Oblig.

[Teancum]

You seriously believe that all atheists are communists and accept the teachings of Mao?

I think you are mistaken about something here that I was trying to say. I said nothing of a universal faith that can be called atheism here, but rather was describing a whole spectrum in various dimensions of atheistic belief.

There isn't "one true atheism"... such as it is. It isn't nearly so neatly organized, and for those slashdot readers who are seemingly offended here about my spouting heresies about atheism should really go back and think about this some more before getting offended here. Obviously I've touched on a nerve on this issue.

I'm not going to continue this thread any more, as it is starting to go back and forth in a series of "yes I am" "no you aren't" type of cycles, but I do think there is some massive misunderstanding here of what I'm saying. I started out by suggesting there were many forms of atheism and named off several fairly common branches of atheistic thought... all of which seemed to bring offense to those claiming to be atheists.

This has been a fun thread, but it also missed the point completely about the existence or non-existence of a soul and if that may be critical somehow to the field of artificial intelligence. The one thing that seems to be a common thread among those who do profess a faith in a higher being is that there is something eternal and infinite in the existence of a "soul" that continues on after your physical death. I was trying to argue here that such a definition of a soul can also be described in strictly mathematical terms that doesn't necessarily even require the existence of a "God", even though many would find it comforting to find a "proof of the soul" through AI research.

Re:Oblig.

[Anonymous Coward]

intelligent as a bug or a small animal

Emotions and language seem very far off None of that is right. To develop his http://en.wikipedia.org/wiki/Psi-Theory
http://en.wikipedia.org/wiki/Dietrich_Dörner developed a simulation with animal like
creatures that are quite intelligent, learn to behave in new situations, learn the "grammar" of things,
learn from experience, have motivation and in fact have emotions. And all of that works today on your average PC.

Re:Oblig.

[mdwh2]

I think you are mistaken about something here that I was trying to say. I said nothing of a universal faith that can be called atheism here, but rather was describing a whole spectrum in various dimensions of atheistic belief.

Atheists have beliefs, sure, this doesn't mean that there is a belief system or religion called "Atheism", nor are these beliefs a form of atheism.

If an atheist believes that communism is a good system, or that chocolate is a tasty food, these are not forms of atheism, anymore than they would be forms of Christianity if a Christian believed them.

This has been a fun thread, but it also missed the point completely about the existence or non-existence of a soul and if that may be critical somehow to the field of artificial intelligence.

Until someone gives an actual definition of "soul", there is no point to miss ("something eternal and infinite, that continues after your physical death, are vague properties of this soul, but they don't tell us what it is). It makes about as much sense as me saying that Duke Nukem Forever can never be released, because it requires some magical but undefined thing that I will call a "wibble", that cannot be reproduced by mere mortal programmers. What is special about intelligence? The only thing I think you could possibly say is that intelligence isn't reproduceable on a classical Turing machine, because of its complexity (in which case, a classical computer couldn't reproduce intelligence, but that doesn't mean we could never build an intelligent machine, e.g., using a quantum computer).

I agree though that belief in a soul - whatever that is - doesn't have to be correlated with belief in God; theoretically an atheist could believe in souls. Or fairies, the easter bunny or Father Christmas.

Re:Oblig.

[Teancum]

Even with the stricter definition that you seem to be applying to religion, it is possible to be an adherent to multiple religious philosophies.

As far as your speaking on the behalf of "most people"... you are treading on some very thin ice by making such a statement. Please don't take offense in what I'm saying, but instead use it as an opportunity to examine just what you are claiming to call a religion.

I'll admit that many individuals to tend to compartmentalize their lives in terms of "work", "school", "home", and "religion", and even have time that they spend at each of these in a nearly exclusive manner. The problem with this is that you aren't being honest with yourself if you practice a religion in this fashion without considering that the religion is a greater part of your being and who you are. Genuine adherents to a religious philosophy will let it "consume" their whole lives and make it a part of their daily lives in all areas. Those who profess a "faith" in a regimented manner, I believe, are simply doing it for some sort of social face saving move and really don't adhere to the doctrines of that religion. I am also asserting here that they aren't being honest in what their true religious compass is trying to tell themselves.

Religion doesn't necessarily require a great religious leader such as Jesus, Buddha, or Muhammad, as apparently you are trying to assert here.

Re:Oblig.

[mdwh2]

And if you can run them awful quicker than nature, then they are all awful approximations...

Think about it, suppose we have a program that is almost on par with human intelligence. Obviously ONE such program is going to need some rather nifty hardware to run.

I presumed human intelligence was the end goal of this idea - for most of the billions of years, brains were much simpler, and for the most part didn't exist at all.

Obviously it would still take vast amounts of computing power, but it wouldn't have to take the same amount of time, as you don't have to wait around years for every organism to physically grow. Nature is pretty much itself an "approximation", with some many random factors and variables - it's not clear that every last detail of where a little rodent walks throughout the years of its life is necessary to reproduce.

Re:Oblig.

[Jon Peterson]

"Language isn't far off at all, we just about have it already"

Actually, we aren't even close. Not only that but we've barely advanced in the last 20 years. Don't get me wrong, we're seeing lots more stuff in industry doing 'NLP' - but what its doing was done in labs years ago. It's a case of industry closing the gap on academia. The academics aren't making much progress.

We can use crude dictionary based algorithms to find noun phrases. We can have a fairly good stab at linking adjectives and adverbs to their nouns and verbs. We can get negation right 80% of the time. That's just parsing simple sentences. And sentences are the easy bit of language. Attempts at context sensitivity are dictionary based and poor.

For instance, what does "it" refer to in the following bits of language?

"It is raining"
"Eureka! I've got it!"

Think about how on earth you'd get a computer to determine the referent of 'it' in each case. NLP is not the route to AI. If anything, when we get further with AI in a couple of centuries, we may be able to make headway on NLP.

Re:Oblig.

[fyngyrz]

Er, yes. :-) All the verbing of a noun is, in essence, is emphasis. Ball. Fast ball. Dogs and cats do this all the time. Want food! Food is not forthcoming. REALLY want food!

What you're really talking about is the level of abstraction; and that's not something they're going to communicate if they're not doing it in the first place.

Re:Oblig.

[seanor]

I completely agree with this assessment. There is in the IT crowd a certain group of people that seem to think that intelligence is digital or can be mocked up by digital manipulation of various unknown systems. Real intelligence, i.e., the ability to know that one is knowing and to reflect upon that process requires a kind of identity that the IT enthusiasts (I was going to say pinheads) completely miss. If this capacity is ever duplicated it will have to be at the quantum level and will involve processes completely unknown today. I suspect that this won't be doable for another 200 or 300 years. The section of the IT movement that focuses on machine intelligence would do well to study metaphysics and particularly ontology and the dynamics of epistemology. Trying to grasp intelligence without studying the conclusions of centuries of thinkers before them is short-sighted and ignorant.

Re:Oblig.

[fyngyrz]

No, I *don't* want "human intelligence, not just intelligence." I just want intelligence. I don't even know why some people are so hung up on the idea of "human" intelligence in the first place. Humans have very mixed records (cough) Bush (cough.)

As for the rest, those are your opinions and you are welcome to them, but I don't share them -- any of them. I don't view AI as "stuck", I don't think human grey matter is the limiting definition of where the method of science can come from (it is, after all, a very simple method), and I know for a *fact* that solutions can arise where understanding is less than complete; been there, done that, wore the surprised look.

But I'll share that pizza with you on The Day, however we get there, and whomever does it, using whatever approach. :)

Re:Oblig.

[JimFive]

Most of us know jack about the algorithms that allow us to catch a baseball in flight, yet we can still do it.

Outrageously incorrect. I've always hated this example because it isn't true, or even useful. Most of us know jack about the physics equations that describe where a baseball will land. All of us who can catch know the algorithm for catching that ball. 0. Repeat 1. Guess where the ball is going to land 2. Move to that spot. 3. Stick out gloved hand 4. Until ball is in glove or ball hits ground. -- JimFive

Re:Oblig.

[JimFive]

You'd think I would have learned to preview by now.

Most of us know jack about the algorithms that allow us to catch a baseball in flight, yet we can still do it.

Outrageously incorrect.
I've always hated this example because it isn't true, or even useful.
Most of us know jack about the physics equations that describe where a baseball will land.
All of us who can catch know the algorithm for catching that ball.

0. Repeat
1. Guess where the ball is going to land
2. Move to that spot.
3. Stick out gloved hand
4. Until ball is in glove or ball hits ground.

--
JimFive

Re:Oblig.

[Khazunga]

You have a very distorted view of science. Science is not dogmatic. Current science is math based. Math lays, not dogmas, but axioms. The distinction is relevant, because axioms can be challenged. In fact, they are. There are whole branches of Math investigating different Math models based on different axiom sets.

Your error in analysis is taking people's dogmatic behaviour at face value. One someone studies string theory for their PhD, they're sure to defend string theory as if it were a dogma. Their behaviour does not make string theory a dogma. It's a theory, with all that it implies. People may behave dogmatically, as we're all emotional animals. Science, however, is not dogmatic. The very core of the scientific process relies on challenging theories.

As for sacred books, I completely miss your point. A sacred book is the book that defines the dogmas for a religion. The Bible, The Bhagavad Gita, The Qur'an are all sacred texts. They all define dogmatic sets. None of the books you mention defines one dogma, let alone a dogmatic set.

Anyhow, unless your point is that science is a religion, all of this science related discussion is moot. From the Concise Oxford Dictionary, a religion is: "Human recognition of superhuman controlling power and especially of a personal God entitled to obedience".

Obviously, it follows that atheism is not a religion.

Re:Oblig.

[spun]

That last link ended with a message that can apply beyond brain simulations: "Size doesn't matter; it's what you put into your model and how you embed it into the environment." Sorry, girls are lying when they tell you that.

Re:Oblig.

[spun]

We won't ever be 'obsoletized.' Right now, in any particular capacity, there is someone out there who is better than you. Does that make you obsolete? No. Whatever you do frees up the AIs to do something even more useful with their capacity. And most of the world's population today does not contribute that much to art or science, yet they are not 'obsolete.' The worth of a human being does not stem solely from their contributions to art and science, but also from their interactions with other human beings.

Re:Oblig.

[Anonymous Coward]

Sounds like a job for BOINC....

Re:Oblig.

[cp.tar]

Er, yes. :-) All the verbing of a noun is, in essence, is emphasis. Ball. Fast ball. Dogs and cats do this all the time. Want food! Food is not forthcoming. REALLY want food!

What you're really talking about is the level of abstraction; and that's not something they're going to communicate if they're not doing it in the first place.

Your examples have nothing to do with verbing of nouns.
And the level of abstraction is a part of the cognitive leap.

Re:Oblig.

[NoOneInParticular]

And seeing that strictly speaking, 'consciousness' is merely a label, and not a scientific term, nor an observed fact, it is not even clear that it exists, let alone that it as a separate entity that can be studied.

Re:Oblig.

[advance512]

Actually, it'll be 1048576 times as powerful as today. But, yeah :)

Re:Oblig.

[Easy2RememberNick]

This was it:
  http://news.bbc.co.uk/2/hi/technology/6600965.stm

The team, from the IBM Almaden Research Lab and the University of Nevada, ran the simulation on a BlueGene L supercomputer that had 4,096 processors, each one of which used 256MB of memory.

Using this machine the researchers created half a virtual mouse brain that had 8,000,000 neurons that had up to 6,300 synapses.

The vast complexity of the simulation meant that it was only run for 10 seconds at a speed ten times slower than real life - the equivalent of one second in a real mouse brain.

Re:Oblig.

[HuguesT]

No he didn't. He ran a simulation of a large bunch of neurons equivalent to a 300mm x 300mm area of flat neuronal tissue, which ran for a month or so. The simulation exhibited some lifelike patterns like alpha-waves, but apart from that, sorry, no thought or philosophy came out, and no real insight about brain structure came out. There is a nice video of the outcome.

He simulated a simple physiological model of a collection of neurons, but not a brain at work. In his own word he could have done the same work with a 1000 neurons simulation, but he wanted to do the large scale one to show it was possible.

He estimates we can maybe simulate a number of neurons similar to that in the brain in real-time from 2016 onwards (just barely though), however, as he writes "many essential details of the anatomy and dynamics of the mammalian nervous system would probably be still unknown."

Re:Oblig.

[jdoeii]

> The simulation exhibited some lifelike patterns like alpha-waves,
> but apart from that, sorry, no thought or philosophy came out

When a baby is born it has all the brain structures, it runs in real time for weeks, demonstrates all kinds of brain waves. But try extracting thought or philosophy out of it :-).
The purpose of Eugene's experiment was to demonstrate one can simulate life-size brain with contemporary hardware.

Brain simulation with modern hardware is just too hard because hardware is not fast enough. Models have to be sufficiently big. They have to be run for a while to allow for learning. Then there has to be many experiments to tweak model parameters and test various conjectures.

We need comparative cephalomics (cephalogy? sapientology?): change model parameters and observe consequences. That's not possible yet. Like comparative genetics was not possible 20 years ago. We are held back by hardware. Software will follow quickly.

Re:Oblig.

[Anonymous Coward]

| The cognitive leap a simple verbing of a noun requires is beyond any other animal.

A bold statement, do you have any prove to back it up?

You must be aware that human intelligence is evolved, and that a lot of animals share characteristics like emotions and feelings; even humor and ethics can be observed.
It is logically and factually impossible to deny this, unless you define intelligence as exactly what it is to be human.

Re:Oblig.

[cp.tar]

| The cognitive leap a simple verbing of a noun requires is beyond any other animal.

A bold statement, do you have any prove to back it up?

You must be aware that human intelligence is evolved, and that a lot of animals share characteristics like emotions and feelings; even humor and ethics can be observed.
It is logically and factually impossible to deny this, unless you define intelligence as exactly what it is to be human.

FWIW, the best definition of intelligence I've encountered is "the ability to correctly solve IQ tests".

That aside, IAALinguist.
IIRC, Tomasello argued — quite convincingly — that one of the things that separates us from the other animals is the cognitive ability to share attention and put ourselves mentally into the role of another. OK, so maybe it's two abilities, I forget.
Verbing a noun — on the cognitive side — requires use of language, which is related to those abilities, and then some: the level of abstraction high enough to abstract an action connected with the noun in question and invent a verb to signify it.
Sintactically and morphologically, it is trivial in languages such as English, but still it is quite a cognitive feat.

And all that is just for very simple word creation — what about inventing completely new words? If you have any evidence of animals devising new signs when the need arises (they've done some very interesting experiments with bees, BTW), I'd be very interested in it.
Until then, though, I'll stick to my current POV.
And BTW, I was a great believer in the language ability of animals too. Before college. Studying language made me realize it was a much more complex phenomenon.

Re:Oblig.

[Anonymous Coward]

| My current software does what I said, but it's too slow on current hardware.

Ha ha ..., so prove it. You can you say.
Put your computer(s) and program in a controlled environment (and we will check for midgets etc.)
We will do the Turing test.
What do you say, tomorrow at M.I.T.?

However, I know the outcome already.

Re:Oblig.

[Teancum]

Perhaps if I have a "distorted view of science" it is because I know professional scientists, have attended scientific workshops and forums, and have even participated in "pure" scientific research.

Real science is something that often isn't pretty, and there are huge egos and reputations on the line by people with often powerful political connections that make things happen.

You have also missed something critical here. I have not said that science is a religion. I said that science, to some individuals, may be for all intents and purposes a religion unto itself. And I suggested that *some* atheists (certainly not all of them) who genuinely reject the idea of a supreme being may be "adherents" to this "religion of science".

As for the Oxford Dictionary definition... I question the validity of that definition. That may be a more or less reasonable definition in many cases, but in terms of practical applications of a moral philosophy that guides your life, this definition is very limiting. I am asserting here that religion can be defined much more broadly, and that your application of explicitly excluding atheism from the definition of a religion is a deliberate attempt to segregate atheism from other religious philosophies... and for a political purpose as well. Of course my (and others who feel the same way I do) attempt to challenge this definition used in this manner is also a political action. That is why this has touched such a raw nerve.

Re:Oblig.

[Khazunga]

Perhaps if I have a "distorted view of science" it is because I know professional scientists, have attended scientific workshops and forums, and have even participated in "pure" scientific research.

Real science is something that often isn't pretty, and there are huge egos and reputations on the line by people with often powerful political connections that make things happen.

Very true. Yet, despite being discussed emotionally, if you look back from some distance, the scientific method has kept its core feature: an always questioning view of the world, including past theories. This is the core of science, and directly contradicts religious behaviour which is always dogmatic and unquestionable.

As for the definition of religion, I dare you to come up with a definition that is capable if including atheism, which is not even an organized social group, without including other social groups that are clearly not religions, like mountain hikers, sports fans or knitting groups.

Re:Oblig.

[Walt Dismal]

I actually appreciate your skepticism, because it helps me analyze and improve my position.

I will indeed go for the Turing competition eventually. But you should read what I said more carefully. I did not claim to be running everything integrated and fully functional today. This is an enormously complex system, and the pieces to be integrated are not conventional code, they are theory implemented in NNs. That is a lot of work. But I have significant pieces right now that do what no one else currently can do, and chunks planned but not implemented from the underlying theory yet, and years more to go within my development. But enough done that it's clear it's all on target. I said a time frame for human-like AI lies within Kurzweil's prediction. And I expect to have major pieces pulled together within two and certainly five years.

Also, it seems you're thinking conventionally. You call the AI a program. That shows you don't have the right model. Although some people run AI routines in software, full strong AI is nothing like conventional software. My AI is not an AI program, a piece of code. It is AI emerging from neural nets, and some new concepts, under the right machine architecture. That's nowhere the same thing as a program in C++ or LISP. Very different. My current software runs low level code that merely runs NNs on a Von Neumann platform, which is simply not optimal for MPP. The intelligence comes from the way the NNs connect, how they learn, what they learn, and how they control the dynamics of their organization. It is a layer above the code and is nothing like conventional programming, and I admit hard to understand.

Re:Oblig.

[Some_Llama]

"We already know that it's possible to contain 100% of real-time human brain functions in a casing 10cm by 10cm by 10cm and weighing under five pounds.
Do you have a source on this by any chance? I'd be interested in reading how we realized that this is possible."

I'd be more interested in this source:
"it's just another personal platform to play Duke Nukem Forever"

Re:Oblig.

[crimson30]

I've been saying that for years now...

I liken it to genetics. We have the technological ability to make a human that lives for 200 years. But we simply don't have the knowledge about the genome to make it happen.

Just because you have the technological ability, doesn't mean you instantly gain full understanding of every application possible.

Re:Oblig.

[Anonymous Coward]

| I actually appreciate your skepticism, because it helps me analyze and improve my position.

I think you would be skeptical if someone else had a similar claim.

| But I have significant pieces right now that do what no one else currently can do, and chunks

You don't know that for sure. Someone could be on track, but isn't telling.

| planned but not implemented from the underlying theory yet, and years more to go within my
| development.

Thats all fine and well. But to be sure of your system you must be able to test it. If you cannot do that your claims are unscientific. And you shouldn't post them.

| But enough done that it's clear it's all on target. I said a time frame for human-like AI lies
| within Kurzweil's prediction. And I expect to have major pieces pulled together within two and
| certainly five years.

If you can test an important part, as you imply, you should be able to demonstrate that.

| Also, it seems you're thinking conventionally.

That could be. But, technically I'm right. Your running a simulation of a neural network on a conventional processor. And thats a program.

| You call the AI a program. That shows you don't have the right model. Although some people run AI
| routines in software, full strong AI is nothing like conventional software. My AI is not an AI
| program, a piece of code. It is AI emerging from neural nets, and some new concepts, under the right
| machine architecture.

I know a lot about neural networks. I worked with neural networks at the university as far back as 1986. I know of the theory and of the limitations.

It is an obvious idea to use neural networks to create an artificial intelligent program. But in practice, the concept is unusable. It's a dead end for a real understanding of consciousness and human intelligence.

Re:Oblig.

[Walt Dismal]

"But in practice, the concept is unusable."

No. Only as practice is currently implemented by others. Not like what I have. Your vision seems to be within the conventional box. In a way analogous to Marvin Minsky shooting down NNs for a long time because of his failure of vision, inability to conceive of other forms of the technology. Think larger and don't make his mistakes. Go read Beiderman's work, for example, on geons in visual perceptual systems. This hints at what properly-architected systems can do. Now extrapolate from this to what much larger systems could do with knowledge. You have NN experience, but you're stuck in the rut of the current paradigms. There is a region between the limited NNs people currently implement, and the massive ones of the brain. My work lies between these two, and I've come up with a new approach that merges connectionist and symbolist. I already have systems that can understand many of the things missing from sterile AI implementations: emotion, human belief systems, even humor. This is how I know the complete implemented architecture will be viable. I have a different approach than the failed visions. The underpinnings are solid, the model is solid.

"But I have significant pieces right now that do what no one else currently can do, and chunks"

"You don't know that for sure. Someone could be on track, but isn't telling."

Sure. But right now, nobody else has AI that understands humor. Not even close. I do. Published papers are completely off the mark, and I'm very confident of leading the effort. Ditto in areas of AI dealing with emotion and many other things required by an AI to understand people. As I said, the OCC standard model is in error and has holes, and they are the gold standard. It's easy for you to say someone could be on track, but really, published work doesn't show this. Who will have the best model and implementation? Mine works very very well, it IS tested. Lots of subtleties yet to be expanded, but it's clearly working. There is competition certainly, but there's no evidence of anyone out there who's pulled together all the things I have. I'm synthesizing across a broad range of fields. There are very few doing that with the same scope and scale and current results. Hence my confidence in my work.

Re:Oblig.

[Plutonite]

That would be a point of argument then, since unfortunately much of the research is focused on the human intellingence part, mostly because that's how the intelligence is defined to begin with. And like I said above, your scientific method will be pretty useless without the creative and abstractive elements provided by the very interesting phenomenon of sentience in the human brain. It is not the grey matter that is inherent to science, it is the constructs provided by the interactions in that grey matter. Stupid humans and smart ones (which have almost identical DNA) are again perfect examples of why this intelligence issue is so damn complicated.

But I agree with you - I am part of the "screw the humans" camp which attempts to guide AI by virtue of the capabilities of formal logic alone.

Re:Oblig.

[NicknamesAreStupid]

Shortly after humans invented 'language', somebody said that "man will never fly." Within the context of that reference, it is true to this day. That primitive was thinking of a man flapping his arms and soaring through the skies. In the same regard, you are right; animals never speak English and conjugate verbs. However, looking outside the box, man does fly in aircraft and animals may have an abstract audio communication system as sophisticated as human language that we simple have not yet deciphered.

Re:Oblig.

[Anonymous Coward]

No. Only as practice is currently implemented by others. Not like what I have.

That could be, simply because it is possible. We are living proof of that.
But as I stated before, if you have prove, you should present it. Then we can all decide if it is valid or not.

Apart from that I didn't explain why I think it isn't practical. In short, to recreate a brain in a set of neural networks it is essential to be able to record that information from a human. This is practically impossible. The current state of neuroscience is not even close to be able to give this information. A few theories and some actual samples are clearly not enough. A measuring device that can read the state of all (or at least a large part of all neural regions) would be required.
This isn't feasible, not even in the long term, I think.

My aim is to truly understand consciousness and (human) intelligence. And as I stated before this cannot practically be achieved from a neural net perspective. Simply because neural nets are at an ultra low level. The only way to go is to create a real understanding on a higher level independent of the actual hardware to implement it (hardware isn't even interesting from this perspective).
And in this respect neuroscience, information science and philosophy does give us more insight (I would like to recommend work of Daniel Dennet and Douglas Hofstadter).

Your vision seems to be within the conventional box. In a way analogous to Marvin Minsky shooting down NNs for a long time ...

I wasn't aware of Marvin Minsky's dislike of neural nets (I read some of his work a wile ago). But as I stated above I think I do understand why. Marvin Minsky is an ultra clever person, so I do like it to be compared with him. But personally I question if I'm in the same league.

Re:Oblig.

[HuguesT]

Yes, quantum mechanics can be simulated, but very inefficiently. That thought requires quantum operations would be a major discovery. At the moment it's speculation at best.

Re:Oblig.

[Walt Dismal]

"to recreate a brain in a set of neural networks it is essential to be able to record that information from a human. This is practically impossible."

Let me clear something up. I haven't claimed anywhere to replicate existing people's brains. I can't stick some science-fiction scanner helmet on a person and make them live forever in an AI. However, I think it's quite possible to simulate people to a usable workable extent by building a model of their personality, traits, belief systems, estimates of what they know. My Groucho model does exactly that, for example. This is useful in making models of people in foreign cultures. For example, if you were to put a Muslim model in a situation and offer it pork and alcohol, the predicted response would be that its cultural beliefs would drive it to refuse the items. By combining cultural rules and the personal beliefs of the modeled person, you can provide some human-like semblance of behavior to some of an AI's behavior.

NNs are merely low-level components in an architecture. Intelligence emerges from much higher-level functions built in layers upon the lower-level functions. An inexact analogy is the OSI 7-Layer model for the Internet. No one layer IS the Internet, the behavior of the Net arises from the operation as a whole.

Re:Oblig.

[Anonymous Coward]

"to recreate a brain in a set of neural networks it is essential to be able to record that information from a human. This is practically impossible."

Let me clear something up. I haven't claimed anywhere to replicate existing people's brains. I can't stick some science-fiction scanner helmet on a person and make them live forever in an AI.

You didn't understand me, and I was a little unclear.
To rephrase: to be able to train a set of neural networks to become a coherent brain, it is necessary to know the basic layout of a human brain. You have to know the main communication channels between brain areas, the basic layout of the neural areas and the structure and connections to these areas from the senses. In short, you have to know the basic 'programming' (or genetic hardwiring) of the brain, for it to be able to learn.
This is not the same as a record of the actual state of a specific human brain. But still extremely hard to do, and impossible by far at this moment. Without it, the neural nets cannot be trained, only chaos will result (comparable to the medieval 'scientists' who tried to create a living human by mixing the elements).

However, I think it's quite possible to simulate people to a usable workable extent by building a model of their personality, traits, belief systems, estimates of what they know. ...

This doesn't make a lot of sense to me, it seems that you claim to train your software in a rule based manner. But this isn't compatible with a neural network. Your example is also suspicious, I don't believe you have actual software that does that.

But time will tell, in about 5 years we know the answer (according to you). And maybe I will contribute too.

Re:Oblig.

[Bobb Sledd]

I would like to talk to you some time about AI in translation software. After studying several languages, I cannot see how we will ever achieve automatic translators without AI. There are too many synonyms for things that you want to say; it is difficult enough to use an automatic translator to say what you meant without becoming garbled.

I believe that an automated translator must first understand the original text, parse it into an abstract idea, and then translate that to the target language (not simply source language to target language). To do that, means it must understand context. To do that, it must understand what subject we are talking about now, and understand attributes of objects in the subject (even though those attributes were not mentioned). And I'm not even bringing up the understanding of the social situation necessary for the right pronouns (such as which form of "you" and "your" you should use).

This is why we still don't have a decent Chatbot. Because a chatbot only responds to likely responses. It has no concept of the current "subject", so it consequently renders back irrelevant statements.

Anyway... just an idea or two.

Re:Oblig.

[Walt Dismal]

You're quite right. I'm working on this problem and will share some views. The basic problem has many layers, so I can't cover it all here, but that's partly why I'm writing whole books about this. All language is tied to the beliefs about the world of the culture using that language. So all words, phrases, and idioms of the language are very much tied to the belief systems governing how the speakers think, as well as use that language. In a conversation between speakers from different cultures, the source must rethink everything he wishes to communicate and recast it using the thought systems of the target language. Simple word-for-word lookup will often fail to properly do the job. So translator software systems must understand world models and the complex beliefs that operate within those models, and map (or create new mappings) from one system to the other system. If you have a culture that counts "one fish, two fish, many fish", a translator must understand how that culture thinks about number and remap your statement into their logic frame. To do this, an AI must know about belief systems. Further, it must recognize what cultures each party lives within. I am writing a whole book about that. It is in part a problem encompassing philosophy, logic systems, psychology, sociology, and anthropology. AI has failed to take this into account properly, which is one reason it has fallen so short dealing with people problems. The kind of engineering mind that handles computing does not usually handle the humanities too well, and vice versa. This is why few have worked on integrating these domains before. Chatbots will not rise to the complexity needed for Turing tests until they can build good world models handling these human factors. I expect to see AIs that you can ask "how do you feel about garlic snails", and they will be able to answer "I don't care for eating snails, I just like pizza." if the AI has been configured for North American culture, as opposed to French.

Re:Oblig.

[Walt Dismal]

to be able to train a set of neural networks to become a coherent brain, it is necessary to know the basic layout of a human brain.

Ah. This where our models differ. I believe it is possible to produce functionality that does what the human mind does without having to duplicate the exact mechanics of the human brain. A simple proof: we can easily write code that does arithmetic. Yet it does not run on a machine that duplicates human neural structure. At a more complex level, we can build machines that handle knowledge and reasoning yet do not (necessarily) use NNs. My approach spans the two kinds of system. It is hard for people to understand how this works. I use NNs and other mechanisms and different architectures than a human brain to produce the same quality of capability. There is no need to know exactly how the brain is wired if you can produce the same functionality other ways. However, the other way can employ some of the means the brain does. Just not all of them.

I am not claiming I have a simple rule-based system nor do I train it that way. What I have is far more complicated. I have a whole architecture using connectionism at the base level but which runs symbolist in effect, at the higher levels. In effect the NN output represent tokens, and at the higher levels the tokens are symbolist knowledge operating within connectionist architectures. That IS what the human brain does. Realize that, at some level, the brain is storing "C-A-T" and working with it. Yet this is totally within NNs. In the human brain, the belief that "Carrots are good for you" gets stored within NNs, although not directly as 'carrots+are+good+you". Rather as links between nets representing the concept of carrot, the concept of good, etc. I am simplifying a bit here, the mechanism is much more complex than that. Way more.

Re:Oblig.

[HuguesT]

Actually a baby has plenty of hard-wired and flexible structure already working fine doing something useful. I'll spare you most of the detais, but babies can breathe, eat, etc. They can barely see but they can hear, smell. they have a working memory, and they can communicate effectively, and not just by crying.

The brain tissue simulation had no particular structure. If they had been able to simulate say memory from neuron collections I would have been very impressed. This was just a simulation of a collection of simulated primitive neurons firing and inhibiting each other with no outcome. We don't know how to set up useful neuron structures.

Re:Oblig.

[jdoeii]

> We don't know how to set up useful neuron structures

In this particular case, yes, but in general I tend to disagree. There were successful simulations in which RNN learned to separate spatio-temporal patterns (that's the basis for intelligence - separation). They can be designed to have memory through (a) dendritic delay lines, (b) short term potentiation of synapses, (c) long-term synaptic plasticity. The (b) can be seen as a short term memory, the (c) is the long term memory. (a) is a working memory on the scale of ~1 second.

Re:Oblig.

[Bobb Sledd]

Yes! I have also thought most of that as well!

And I cannot believe that you and I are the only ones who have thought that... but why then is there such a lack of visible AI that takes any of that into consideration?

The other problem that also exists is the lack of the machine's ability to imagine a next logical step in a process, when the end result is not really known. Or; to realize facts that aren't explicitly given. Why is that important? Because when people speak, they do not include all the necessary facts to deduce the meaning. Perhaps the fact has already been stated in a previous sentence or paragraph. Or perhaps it's just human common knowledge. But for the correct translation, that fact might be necessary for correct translation.

I think of a time when I was trying to translate the rules for Yahtzee into Russian. I didn't even get close! (According to my real-live human translator). The translation system kept translating "points" roughly to "glasses" (spectacles). Even though I knew what it was doing, I could not come up with another word to describe the concept of "points" as it pertained to "game."

Real AI should have known that the subject was "game" and not "seeing" and been able to deduce the correct word.

And yes, culture was also a big problem -- it is a big problem in telling jokes in another language. Perhaps if you have someone who can translate well, they *might* possibly be able to translate your joke, but not usually well.

Or take a play on words: "Time flies like an arrow. Fruit flies like a banana." In Russian, it loses all meaning.

Anyway, this is a subject of much interest to me. And I've often wondered if there were any kind of database that could assist in an AI's learning, and what it might look like (at least specifically to language).

Cheers,

Re:Oblig.

[Walt Dismal]

Others have thought of it, but linguists have not bridged this to AI concepts needed.

I believe that all SIs (synthetic intelligences, the term I use for human-like intelligences instead of AI, as AI denotes perhaps a sterile intelligence) must utilize what I call culture bases. A culture base is the system of beliefs of a particular cultural group. It is usually particular to that group. Thus, Catholics have a Catholic culture base, mathematicians have a culture base related to mathematics, etc. To understand a Catholic, to understand a mathematician, you must have knowledge of their culture bases even if you do not have and use their bases in your own life. Thus, I must know of the religious beliefs of Catholics - or Muslims! - to understand how they view life, even if I do not share their beliefs. This is a key factor in translation too.

A difference between animals and humans is that humans can pass their culture bases on to new generations through language and other means, but animals may be restricted to mirroring and learning activities. However, some activities may over time become genetically embedded in the sense of animals evolving with certain behavior selected for. For example, mating rituals in birds are clearly genetically embedded. So do humans have some embedded cultural behavior? I am exploring the extent of that, I'd say yes, some. But what is the range of this? A good topic for research for some doctoral candidate. An example, women respond to big-eyed small creatures (infants) with what is probably a genetically-determined nurturing stance.

Coming back to language, a part of a child's learning a language involves the process of learning the cultures of the parents. So the databases, as there are multiple ones, of an SI will be formed as it learns the worldview and activities of the cultures it is intended to deal with. Since there are substantial numbers of cultures and subcultures, the problem is complex. An Si will have to handle a mix of bases, and reason through the inevitble collisions of views, of worldframes, between cultures. In my work, I'm analyzing the nature of cultural knowledge needed for an SI to understand humans. Storing the culture bases in a NN architecture involves massive nets, which is why I need a new class of computers to run things on. Current experiments using clustering can only implement limited numbers of neurons (see the Darwin machine efforts). I need millions of neurons to handle the data needed for a moderate SI.

Factors to think about when designing an SI for the Turing competition: how could it handle a question about "what games did you like as a child?"; "What flavors of ice cream do you like, and why?" I believe that SIs must come to have beliefs, values, and even tastes like people do, or they will never be able to understand how humans think. You can see the obvious relevance to language translation here.

Re:Oblig.

[ralphbecket]

A late reply, but, IMHO, the problem with "Society of Mind" (and other books in that vein) is that it's basically just an appealing metaphor rather than a concrete guide. You read it, you think you have a handle on this idea of "mind", then find yourself floundering when you sit down to write an AI program based on the ideas in the book. This is what happened to me and, judging by the complete lack of success elsewhere, I assume it's what's happened to everybody else. I even asked Minsky at a conference whether he'd written any programs based on his "Society of Mind" idea; he responded that, being a mathematician, he didn't do implementation work.

Re:Oblig.

[ralphbecket]

Another late reply - sorry! This is exactly the problem. You could fairly define AI to be "all the computational problems we have no idea how to tackle using existing science and engineering". A large part of any AI PhD is defining precisely what the problem is that you want to tackle. By the time you've done that (no mean feat), you realise that barring a flash of genius you will only be able to take baby steps towards a solution.

As to why I got into AI: it was because I read way too much science fiction :-) I still think it's a worthy goal, but these days I suspect the most fruitful route lies in building computational models of psychological theories rather than the more traditional theorem-prover/neural-net-in-a-box approach. Of course, the psychologists are probably doing a great deal of hand waving in their own theories.

Re:Oblig.

[dubl-u]

You read it, you think you have a handle on this idea of "mind", then find yourself floundering when you sit down to write an AI program based on the ideas in the book.

That's a totally reasonable answer, and I can see why you're frustrated. But I feel like we're decades away from being ready to publish "Society of Mind: The Coding Workbook".

Again, I'm a professional developer, so I know a lot about making software go, and almost zero about what AI has been up to since I gave up on it in the early 90's. But when I look around for the biggest real-world success, I see the Roomba. That seems to validate Brooks's assertion that the way to get to good AI is to start with bodies and give them something to do. And those things seem to be roughly as smart as a mosquito, which isn't a mind that needs a lot of designing.

I wouldn't expect "Society of Mind" to be useful in a practical sense until you're more in the range of mice than bugs. So that it isn't practically useful yet doesn't strike me as a big strike against it. As you said, he's not an implementation guy, and isn't recommending it as an implementation book. Until we've built a few dog-class minds, we won't be in a good position to judge his theories.

Well I'm not holding my breath

[Anonymous Coward]

That will be 21 years until we get the first AI first post

in the meantime, humans will continue to win the frost post battle

am I right, or am I right?

Re:Well I'm not holding my breath

[2.7182]

Yes, I remember well my youth, reading Goedel Escher Bach and Winograd, etc., thinking that the next scientific revolution was coming. Things never got any better than Eliza. Now as a hard scientist, I strongly feel that the problem is far far off.

Re:Well I'm not holding my breath

[Hal_Porter]

The 'people' getting first post haven't been human for years.

Re:Well I'm not holding my breath

[Anonymous Coward]

Things never got any better than Eliza.

Dude, that was 40 years ago. You need to talk to someone to help you get over her.

Re:Well I'm not holding my breath

[BobThePig]

I'm sure the first thing the artificial intelligence will tell us is that it will have nuclear fusion ready as an energy source in 50 years.

Re:Well I'm not holding my breath

[Anonymous Coward]

Who made Ray Kurzweil an AI? :)

Re:Well I'm not holding my breath

[Anonymous Coward]

There is one sentence one of my CS professors uses at least every class: You can program something if and only if you understand it.

No chance

[Kjella]

I mean it could happpen but this is so far from the current state of the art, I think we're talking 50-100 years forward in time. We have the brute powers of computers but nowhere near the sophistication in software or neural interfaces to do anything like this.

Re:No chance

[styrotech]

I think he's correct, but the summary stated it the wrong way. It should've said:

"based on current trends, human intelligence will reach the level of a computer by 2029"

Re:No chance

[Jasin Natael]

Who said our own brains had any real sophistication? For all we still know, it's just so massively redundant and generalized that 90% of our brain is malfunctioning at any given moment, and yet we still have the illusion of continuity.

I'm of the opinion that a great deal of our behavior comes from being hardwired to expect X% of the world to make sense, and when the correlation of input/output from any set of cells in our brain fall outside the "acceptable" range, that group of neurons is modified or reset. I also wouldn't be surprised to find out that emotions are simply chemical changes to tighten or relax the criteria for what's in the "acceptable" range. I know this is kind of out there as far as models of cognition go, but there's no real evidence that the brain is sophisticated and elegant from an engineering standpoint. From an artistic, look-at-the-pretty-patterns angle, it might be amazing, but for such observations as those, beauty is in the eye of the beholder.

We may yet find that the only elegance and sophistication is in choosing X well.

Hrmmmm

[BWJones]

I'll be meeting with Kurzweil in April.... Speaking as a neuroscientist who is doing complex neural reconstructions, I think he's off his timeline by at least two decades. Note that we (scientists) have yet to really reconstruct an actual neural system outside of an invertebrate and are finding that the model diagrams grossly under-predict the actual complexity present.

Re:Hrmmmm

[Anonymous Coward]

I think he's off by a decimal place.

now - ai can beat a human at chess. human designed, and setup the game.

20 years from now - ai can autonomously walk up to a nearby human ask them to play chess, if the invitation is not received well, the robot could make a convincing case as to why the human should change their mind, the robot could setup the board, and initiate the game. In the middle of the game, the ai, would have no way to predict, react to, or analyze after the fact an occurrence of unexpected human behavior in the form of violence, humor, insanity, irrational requests, casual misinformation, or conversation.

200 years from now, most of the above problems will be CRUDELY solved, but still polish will be lacking. ai will not be capable of higher abstract imagination.

we're about 500-1000 years from data's head. several thousand from his head and body.

the last 10% of the job takes 90% of the time.

e

Re:Hrmmmm

[DynaSoar]

And as a cognitive neuroscientist, I say he's off the mark entirely. As per Minsky, a fish swims under water; would you say a submarine swims?

What exactly is the "level of humans"? Passing the Turing test? (Fatally flawed because it's not double blind, btw.) Part of human intelligence includes affective input; are we to expect intelligence to be like human intelligence because it includes artificial emotions, or are we supposed to accept a new definition of intelligence without affective input? Surely they're not going to wave the "consciousness" flag. Well, Kurzweil might. Venter might follow that flag because he doesn't know better and he's as big a media hog as Kurzweil.

I think it's a silly pursuit. Why hobble a perfectly good computer by making it pretend to be something that runs on an entirely different basis? We should concentrate on making computers be the best computers and leave being human to the billions of us who do it without massive hardware.

Re:Hrmmmm

[timeOday]

Please do not take this personally, but I don't think neuroscience is particularly important to AI. Yes, biology is horribly complex. But airplanes surpassed birds long ago, even though airplanes are much simpler and not particularly bio-inspired. Granted, birds still surpass airplanes in a few important ways (they forage for energy, procreate, and are self-healing... far beyond what we can fabricate in those respects) but airplanes sure are useful anyways. I don't think human-identical AI would have much use anyways, since it would have the same neuroses and demand all the same rights that make humans such a pain to work with.

Re:Hrmmmm

[Anonymous Coward]

Your so full of shit. Anyone with anything going doesn't post at Slashdot.

Re:Hrmmmm

[LurkerXXX]

What aircraft corner as fast as barn swallows?

There are still many things we can learn from biology that can be translated to machines. The translations don't have to be 1:1 for us to make use of them. The way birds as well as insects make use of different shapes in surfaces during wing beats have translated into changes in some aricraft designs. They weren't directly incorporated the same way, but they taught us important lessons that we could then implement in different ways but with a similar outcome.

I think Neuroscience does have a lot to teach us about how to do AI.

Re:Hrmmmm

[podperson]

I also think that we're unlikely to equal human intelligence except as a curiosity long after we've obtained the necessary technology. Instead, we'll produce AIs with wildly different abilities from humans (far better in some things, such as arithmetic, or remembering large slabs of data, and probably worse in others). Calibrating an AI to be "equal" to a human will be a completely separate and not especially useful endeavor, and it will be something tinkerers do later.

And I suspect that the necessary insights to produce human-like intelligence aren't going to be around for some time. We still have only a foggy idea of how a lot of human intelligence works in the existing hardware.

Re:Hrmmmm

[Anonymous Coward]

I think what he meant was that by making dumbass statements like this, the National Academy of sciences is *lowering* the level of human intelligence. More statements like this over the next twenty years will actually prove him right.

Re:Hrmmmm

[Flicker]

How many barn swallows can fly at 40,000 ft? Just what are you comparing?

Re:Hrmmmm

[Jeff DeMaagd]

We should concentrate on making computers be the best computers and leave being human to the billions of us who do it without massive hardware.

The thing is, Kurzweil is trying to achieve immortality, which is pretty much predicated on the ability to simulate his brain. I don't know if that's coloring his predictions or not, and it really doesn't say anything about whether there can be a machine that can do a full scan of an entire human brain. I don't know if he'll live that long. He'll be over 80 years old at that time, and to be frank, I don't think he looks like a healthy 60 years old now, despite his voracious vitamin intake.

Re:Hrmmmm

[grasshoppa]

I think it's meant as more a side effect of our work in trying to use nature to further our own machines. No point in reinventing the wheel when we have a perfectly good device behind our own eyes capable of things that take huge and expensive computers hours to do what we can do in a few seconds, and that we're more accurate. Image recognition? Physical movements for robots?

Who knows what this line of research will yield? It may yield "aware" computers, but more likely it'll further the technology beyond anything we could achieve on our own in the same time period.

Re:Hrmmmm

[Anonymous Coward]

Cool. So that means we can enjoy some AI that can actually provide some challenge on any Videogames for like 9 years as the world is going to end anyways at 03:14:08 UTC on 19 January 2038.

Re:Hrmmmm

[Peter Cooper]

I think you heavily underestimate human progress. 500-1000 years to "Data's head?" 500-1000 years ago life was pretty much unrecognizable for those of us in the West today. Technology has not only flourished in that time, but effectively been created (even 400 years ago we had no automation, no motorized transport, no significant infrastructure, no electricity, and Galileo was arguing with the Pope about the Earth being flat!)

If you compare the last 500 years with the last 50 years, say, the amount of progress that has occurred in that mere tenth is significantly higher than in the other 90%. There's plenty of reason to believe that development and progress will continue to advance at even faster rates, meaning that technology in 50 years' time is going to seem not only alien to a human of 2008, but rather unfathomable.

Re:Hrmmmm

[0100010001010011]

Now: A Computer is fast enough to computer all the possibly plays in a game and pick the best solution.

There are computers which can beat humans at Poker (which has less possibly combinations than say chess) but I wouldn't say that a computer knows how to bluff or read a bluff.

Brute force is not AI.

Re:Hrmmmm

[Anonymous Coward]

Speaking as computer scientist, I say you're speaking bollocks.

Re:Hrmmmm

[Anonymous Coward]

>What aircraft corner as fast as barn swallows?

The Suhoi. At least on a per-pound/roll rate or per-pound/turn-rate basis.

Re:Hrmmmm

[Daffy Duck]

Galileo was arguing with the Pope about the Earth being flat!

2200 years ago, Eratosthenes not only knew the earth was round, he measured its circumference. Accurate to either 1% or 17% depending on who you ask. Still, "off by 17 percent" is a lot better than "off by infinity percent because everyone knows the earth is flat, numbskull".

Re:Hrmmmm

[Daffy Duck]

Maybe that's what Kurzweil is getting at: by the year 2029, AI will have achieved human-level abilities to make grossly inaccurate predictions of the future of AI.

Re:Hrmmmm

[Anonymous Coward]

What aircraft corner as fast as barn swallows?

Is that a laden or unladen swallow?

Re:Hrmmmm

[Anonymous Coward]

Anyone with anything going doesn't post at Slashdot

I can testify to the veracity of that statement.

Re:Hrmmmm

[teh moges]

What does a human do to read a bluff? He observes his opponent, takes inputs such as bet size and heart rate, applies them to known patterns of bluffers and looks for a match. Sure a human does this without realizing, but little of how this happens is a mystery. Also, how do humans bluff? They just bet at a negative EV play*, and bluffing properly is a matter of knowing the probability that the opponent will call. I am researching applying AI to poker (look out in June for a lot of high quality research from the AAAI Computer Poker Championship) and this sort of argument, "Computers can't bluff, they just run numbers", is both understating what has been achieved in AI in this field and also overstates what humans do. Yes, computer programs aren't quite up to the standard of world class players (Limit poker has achieved this, but not No-Limit), but this game has only a couple of years to go before this milestone is reached. I predict that by the end of the year, we will have high quality bots that can beat 99% of players, and by the end of 2010 No Limit Texas will be a computer dominated game.

The only thing that humans do that AI doesn't (well) is automatically follow a few paths, rather then look at the whole picture. As an example, it has been shown (sorry no reference right now) that some chess grandmasters look only at a couple of moves and then calculate all the possible combinations from there rather then examine every possible move. This drastically speeds up the calculation, however it does miss moves that could be considered the "best". So while this act of "feeling" which is the best move is a good approximation done by humans, it isn't an optimal or maximal play.

As for the article, I don't agree with all of what he says (the idea of nanobots doing what Kurzweil says scares me and I doubt it will be legal to do this), but I do agree with the 2029 prediction, that is if proper resources are given to that particular problem. Replicating humans is a goal in AI for some researchers, but not all of them. Personally, I couldn't care less if there exists a robot that perfectly resembles a human, as long as there are intelligent computers systems that can do the problems that humans find hard (such as finding patterns in very large sets of data or solving complex mathematical equations).

*Technically, it isn't a low EV play if there is a high probability of the opponent folding. In which case, playing the highest EV play naturally involves bluffing if it can be assumed that the opponent will fold to a bet.

Re:Hrmmmm

[Anonymous Coward]

I'll be meeting with Kurzweil in April....

I have a favor to ask. Could you please - very politely, of course - ask him to knock this the fuck off? There's enough bullshit in the world as it is, we really don't need people actively manufacturing it just to sell a few books.

Oh, also, could you ask him to pass the same along to Craig Venter? I heard they are on some kind of bullshit-oriented committee together.

Re:Hrmmmm

[glwtta]

Please do not take this personally, but I don't think neuroscience is particularly important to AI.

Sure, If we restrict the argument to just the ability to fly, I agree that human plane design has surpassed birds.

My question is: if you take someone who's absolutely baffled by how birds manage to fly, would that person be able to design a functional airplane?

If you are trying to construct a system of some kind, wouldn't trying to understand the only example of such a system that you have, be a good place to start?

Re:Hrmmmm

[Anonymous Coward]

One thing that is sure about Ray is that he, like Al Gore and many other celebrities, loves to be at the center of attention. Ray's self image is that of wanting to be perceived as a fortune teller and in essence a know-it-all. Among other things, Ray thinks he's going to live to 150 years old because he drinks organic vegetable juices every day. Well guess what Ray? Steve Jobs, a wealthy man, has been a vegetarian for years and has had a private chef preparing his meals for years and Mr. Jobs also ended up with a form of neuroendocrine pancreatic cancer (thank "god" Steve has survived this cancer). The point being --> Ray is not "god" and I didn't need any nanobot intelligence to help me understand Ray's self image!

Re:Hrmmmm

[veraguth]

This is in the vicinity of what bugs me about Kurzweil. There's some linguistic sleight of mouth going on when he makes these sweeping statements about machines surpassing humans in intelligence. Sometimes I think that all a machine would need to do to match Ray in intelligence is sit around all day drinking its own kool-aid. I know he's really smart and there's a lot to his argument that technology is advancing at an exponential rate and now in a "fast climb" but none of that makes him a qualified diviner of the future.

Your comparison of airplanes to birds shares some similarities with what Ray does. It chooses some particular advantage the machine has and then projects it onto the whole organism. Airplanes surpassed birds? I can think of lots of ways they haven't, not the least of which is they really suck at landing in trees.

Re:Hrmmmm

[hitmark]

the problem of aircraft corner isnt one of the aircraft, but the pilot.

from what i undertand, the pilot cant take much more then 9-10G, but a airframe can be designed to take more.

also, i think a more valid comparison would be between a helicopter and a barn swallow, that is if im guessing how said bird does it correctly.

Re:Hrmmmm

[glwtta]

Passing the Turing test? (Fatally flawed because it's not double blind, btw.)

The Turing test is not "flawed" - it's a thought experiment on the nature of intelligence. I don't think anyone would seriously claim that passing any specific Turing test would be sufficient to prove that a program is intelligent.

Though I see no reason why it couldn't be blinded properly: take 30 people and 30 computers, randomly assign them to 15 chat rooms of 4, poll them about their conversation partners at the end, and then resolve their IDs to human/non-human. Of course the humans will know they are human, but I don't see how that will make a difference; after all, the computers are tested for their ability to mimic a human who knows that they're participating in a Turing test.

Now, the Meta-Turing Test, on the other hand, I think is bullet-proof: an entity can be considered as intelligent if it creates and applies intelligence tests to subjects of its own creation. :)

Re:Hrmmmm

[Elbowgeek]

A computer could indeed predict human behaviour, as irrational as it may sometimes seem. This is because we are animals, and the computer only need be able to detect which of our animal instincts are at play at any given time. However humans have evolved ways of expressing the standard animal instincts in relatively sophisticated ways compared to other animules, so that does get a bit tricky and often very subtle.

The other aspect of intelligence is harnessing a lifetime's worth of sensory experience which interfaces with our animal instincts and processing it all in real time. But the best computer for all of this is an analog computer, which has inherently greater granularity.

Re:Hrmmmm

[Gogogoch]

African or European?

Re:Hrmmmm

[risk one]

you're misquoting Edsger Dijkstra. He said: "The question of whether a computer can think is no more interesting than the question of whether a submarine can swim". I'm not sure Minsky would agree.

The way I interpret Dijkstra here, is that he meant when a submarine starts to look sufficiently like a fish, we will call its actions 'swimming'. When it has the exact same range and 'functional ability' as a fish, but moves by spinning its rotors, we don't call it swimming. Thus the human criterion for intelligence (according to Dijkstra) is weighted to much by how the intelligent machine does what it can do, rather than the actual range of what it can do. It needs to do it like we do it, rather than just get the same result. I think that line has blurred since Dijkstra said that, and I can say (as a student in AI) that at least these days, researchers in plain AI look only at what humans can do, and try to replicate the results.

That's largely because everybody realizes that we don't have the hardware to do it the same way humans do. Kurtzweil seems to think we'll have it soon, I seriously doubt that in twenty years we'll have something of similar complexity to 20 billion neurons, and even if we do, we'll have no way to configure the linking and weights of those neurons (or whatever it'll be). It's just too complex, and we don't have billions of years of evolution to find the right parameters for us. We need a shortcut, and not enough of that shortcut has been found yet. There might be a breakthrough between now and 2029, but I don't think that would validate Kurtzweil's prediction.

your comment that "We should concentrate on making computers be the best computers... ", confuses me to no end. Computers can perform thousands of wonderful tasks. They're inching closer to mastering language (aspects of it at least), they're recognizing visual input, making diagnoses, controlling complex actuators. Every step along the way to humanoid AI has given us incredibly useful technology. Besides that, the models that the search for strong AI gives us have given cognitive psychologists actual working models of the way the brain works, which can help us to understand things like schizophrenia. And that's a field that's only just begun. This research is one of the most fertile areas there is. Your statement seems to suggest a misunderstanding of what AI actually does.

Re:Hrmmmm

[niewiap]

I totally agree. Firstly, replicating human intelligence in 20 years is a totally unrealistic goal. We are not even at the level of understanding the functioning of the nervous system of a worm with several dozens of neurons, let alone a constantly changing multimillion-cell human brain. I can pretty much picture computers that will behave more or less like a 6-month old human baby in 20 years, but it will be at least another couple of centuries before I can meaningfully discuss the literary style of James Joyce with a machine. Secondly, why the hell bother? Computers possessing human intelligence will inherit human flaws (nondeterminism, emotionality, pursuit of own goals, reliance on intuition instead of computation) and it is precisely these flaws we want to avoid in the so called "artificial intelligence". The main goal of AI nowadays should be to help us make sense of all that data flood we have to deal with. Instead of chasing the golden grail of emulating humans we should try to find good algorithms which will search available resources for relevant data based on human sentence-like queries. I think THAT, and not artificial humans or intelligent nanobots, is a realistic goal for the next two decades.

Re:Hrmmmm

[a whoabot]

How much does reading (as in poker reading, not actually measuring, of course ) the heart-rate of your opponent actually help? I know when I play poker I never try to read another player like that, but rather I concentrate wholly on how bets and cards are and have been in order to determine whether they are bluffing. And then I doubt anyone could read me by trying to figure how "nervous" I am: I'm pretty much the same level of nervousness the entire game: because I don't see any hand as being less important than another, unless I'm sitting out I guess. I'd imagine any other serious player sees it the same way. The one exception would be when I'm already all-in with an opponent: If I have an unbeatable hand then I won't be nervous whereas otherwise I may be, so if you're a third-party and considering whether to join this all-in, then perhaps you could read me and gain a benefit for yourself. But even then, I would be concentrating hard on just that factor, and so become nervous to maintain my regular game nervosity and thereby achieve it.

Re:Hrmmmm

[Tablizer]

Note that we (scientists) have yet to really reconstruct an actual neural system outside of an invertebrate

Octopuses and some cuttlefish are probably more intelligent than many smaller mammals. You are being spine-centric.

Re:Hrmmmm

[nguy]

Why hobble a perfectly good computer by making it pretend to be something that runs on an entirely different basis?

Because that's how one shows that a model actually works.

As per Minsky, a fish swims under water; would you say a submarine swims?

You tell me. People are building models of fish that swim in order to understand swimming better. And it's the same with brains and thinking. If some useful engineering comes out of it, all the better.

Re:Hrmmmm

[joshuaobrien]

"The question of whether machines can think is about as relevant as the question of whether submarines can swim." -- Edsger Dijkstra

Re:Hrmmmm

[Evil Pete]

How many planes can get off the runway powered only by some seeds, and self repair? You underestimate the level of sophistication of biological systems, there is no need for birds to fly at 40,000 feet.

As for AI. Well we have been goddam long time waiting. The more I learn about the brain and neural systems the more primitive our technology looks. And I don't really understand what the fuss is about nanobots, we already have them ... they're bacteria. Just grab some, re-engineer them to our liking. Problem solved.

Re:Hrmmmm

[ppanon]

the last 10% of the job takes 90% of the time.

If you've done the job right, 5 years after you've designed "data's head", the next generation design probably will run 4 times faster, and 5 years after that, 16 times faster. So within 10 years, you'll have a Data that can compress that "90% of the time" into 6% of the time and design his own body in a few years.

Re:Hrmmmm

[isomeme]

Simply put, machines capable of emotional understanding, and other modes of human consciousness, will make better slaves. A machine that understands my tastes, my sense of humor, my personal history, my moods, and the changing directions of my curiosity can better respond to (and anticipate) my needs. Since my needs are not rigidly logical, a rigidly logical mechanism can't meet them.

Of course, one runs up against the question of how one distinguishes between simulating and having emotions. If my robot slave appears to be upset by my disapproval of its actions, is it really upset, or is it doing a convincing job of simulating being upset to fit in better with my style of communication?

At some threshold of complexity, it becomes difficult to answer this question crisply. After all, an angry human is just a happy human with a different arrangement of brain chemicals; one can describe the difference in purely mechanistic terms, in theory -- though of course we can't yet do so in detail in practice.

Asimov's robot stories asked and answered this question in many different ways. In the end I think we'll have to decide that anything that seems conscious is conscious. But I imagine it will be a long and bumpy road between here and there.

Re:Hrmmmm

[lazy genes]

Nice one Dyna, Consciousness is more complex than anyone could imagine.At one level we are the product of 5 billion years of evolution mainly just dealing with a changing enviornment. So it seems to me that consciousness must be atleast 30 billion years old. It seems to be past the quantum level.I wouldnt admit that a god is responsible for it, but sometimes i wish it were true.

Re:Hrmmmm

[sabernet]

To quote xkcd.com

"Turing test extra credit: Convince the examiner he's a computer."

Re:Hrmmmm

[Anonymous Coward]

An African or European barn swallow?

Re:Hrmmmm

[renoX]

>Passing the Turing test? (Fatally flawed because it's not double blind, btw.)

Uh? A 'double blind' experiment is when both the patient and the doctor doesn't know which pill are the effective one. For the Turing test, the interviewer can be alone discussing with the maybe-computer, and if there is really the need to have someone else monitoring the discussion, there is *no reason* why this guy should know whether it's a computer or a human at the other end.

So double blind Turing tests are possible, please stop spewing falsities or explain your reasoning.

As for the rest:
-"artificial emotions" do we have happy or sad neurons? No, so are our emotions artificial?
No of course, so there's no reason why computers couldn't have emotions..
-"making computers be the best computers", the best is also the one who is able to do *anything*, so strong AI is interesting, even if it'll take many more years than Minsky crackpot predictions..

Re:Hrmmmm

[wizardforce]

How many planes can get off the runway powered only by some seeds, and self repair?

self-repair is on its way [materials that can do this are already in development] but...

there is no need for birds to fly at 40,000 feet.

there is no need for airplanes to be powered by an inferior fuel such as seeds either. btw, do you know any birds that use gasoline rather than seeds? can a bird haul thousands of pounds of cargo? at mach 1? birds are interesting but don't think for a moment that their "design" is any better- in fact it's in many ways a lot worse than anything we've come up with so far. we can alter bird genes to allow them to metabolize other fuels [even hydrocarbons] if we wanted, maybe even strengthen their bones etc. but there probably isn't going to be a bird cruising at 400 mph for hours at a time running off jet fuel... ever. planes on the other hand, can be scaled from commercial jet size down to bird size with a corresponding increase in maneuverability at or better than birds. new materials can self-heal using liquid monomers and catalysts imbedded in the material as well as shape memory alloys and plastics that can be bent into absurd shapes while being simple to return them back to their original shape with heat/electric fields. know any birds that can essentially have their bones broken and actually contort themselves in such a way that the bone is healed? no? oops. I guess birds aren't so great after all.

Re:Hrmmmm

[4D6963]

How many barn swallows can fly at 40,000 ft?

Or better yet, how many barn swallows flying at 460 mph could corner as fast as an aircraft flying at 460 mph.

Re:Hrmmmm

[bug1]

As someone interested in philosophy i agree hes way off, you need to understand something before you can build it.

Before creating AI there should be some consensus as to "What is AI", "What is the mind", "What is consciousness", (and as you hint at) "What does it mean to be human" and lots of other fundamental questions.

The current problems of implementing AI cant be solved using computers.

We could consider current computers to posses a level of intelligence now if we wanted to (eg deep blue), it just word games really.

If you teach a monkey to behave like a human, is it human ?

Re:Hrmmmm

[Anonymous Coward]

we're about 500-1000 years from data's head. several thousand from his head and body. So, Asimov got the timescale right after all.

Re:Hrmmmm

[Anonymous Coward]

Obviously Slashdot posters are AIs that are not completely developed yet. Come on now, two references to swallows and nobody asked if it was about European or African swallows ?

Re:Hrmmmm

[fireboy1919]

What aircraft corner as fast as barn swallows?

Most aircraft designs can, accepting for inertia (i.e., if you make a model of an airplane the size of a swallow, and then try to fly it around a corner, it'll do what the swallow can; a 747 sized swallow wouldn't be able to go around corners like a tiny one).

While we're on this, though, a better comparison would be hummingbirds vs. tiny UAVs. I'm pretty sure the hummingbirds are doing a better job not hitting things as they go about their sugar gathering missions than the UAVs are doing whatever they do.

Re:Hrmmmm

[Fatalis]

What aircraft corner as fast as barn swallows?

I'd say these do.

Re:Hrmmmm

[j_sp_r]

Wingtips..... There is a lot to learn from birds, but you miss the point.

Re:Hrmmmm

[ciggieposeur]

Granted, birds still surpass airplanes in a few important ways (they forage for energy, procreate, and are self-healing... far beyond what we can fabricate in those respects) but airplanes sure are useful anyways.

I think you've found a fantastic analogy for the opposite point: computers today are much more like airplanes than they are like birds, i.e. once a computer is set up for a long calculation ("get from here to there") it can do it far better than anything else in the world, but it will not ever be able to independently set up many novel calculations (rapid maneuvers) in order to solve a larger yet less well-defined problem ("find food"). We can get vastly powerful idiot savant type AI pretty much now, which is honestly a lot more useful than true hard AI anyway. But I think that hard AI requires a real understanding of how ape-like intelligence works.

When I was in IT, I used to think that hard AI would spring relatively quickly once there were enough transistors available on one silicon wafer to correspond to the number of neurons/synapses. Now that I'm in chemical engineering, I doubt that hard AI will arrive in my lifetime because I know that the real human body is loosely governed by *millions* of non-linear ordinary differential equations, each requiring parameters that must be measured experimentally, and all of them together requiring far more hardware to simulate than we've got today. Just simulating a relatively few biological pathways in liver cells takes modern-day supercomputers between days and weeks. The ODEs can be often simplified, but only after understanding what they mean physically and making appropriate approximations, and it's that step (understanding) that is the bottleneck in understanding why animal intelligence works physically.

Finally, I think in the end we are far more likely to get artificial bodies for humans before we get to artificial minds for computers.

Re:Hrmmmm

[jrumney]

Not to mention that artificial intelligence at human level has been 20 years away for at least 40 years now, along with flying cars and other such contraptions of science fiction.

Re:Hrmmmm

[Anonymous Coward]

Why hobble a perfectly good computer by making it pretend to be something that runs on an entirely different basis? For the money in it? There are a whole ton of applications that require "human level intelligence" (ranging from assembly line work to mining) that could be better done with machines, and cheaper in the long run. What major corporation wouldn't want to replace their semi-skilled labor with machines, if only they were smart enough to be able to handle and adapt to such situations.
Or consider the market for "virtual friends" or (more morbidly) "virtual ghosts". If a loved one of yours passed away, but for $20,000 you could have them "scanned" before death and be able to continue to at least talk with them after they've gone, wouldn't at least some people consider it?
Lets face it - there are all sorts of applications for "human level intelligence" that we can't even begin to imagine. Maybe we need a machine with HLI to start imaging such uses...

Re:Hrmmmm

[Anonymous Coward]

My dad is better than your dad!!!

holy crap are you retarded.

Re:Hrmmmm

[nnnneedles]

As a computer programmer, I say he needs a good anti-psychotic.

Re:Hrmmmm

[DigitalSkyline]

"I predict that by the end of the year, we will have high quality bots that can beat 99% of players, and by the end of 2010 No Limit Texas will be a computer dominated game. " ... and the cow jumped over the moon. I'd take that bet. Hold 'em is not chess and it's certainly not math. You might as well write the program to make random plays, because unless the bot has the hand won (within the top 2 or 3 possible winning hands), no amount of math (which in the end, is all that computers really do), will make a bot beat an opponent of even average skill. As a programmer who actually wins at poker, I find this entire thread ridiculous to the point of absurdity. Poker is more than the chips, more than the cards. You can not quantify luck, nor can you predict with any certainty the actions of a player in any given hand. Any player can make the mathematically correct move 100% of the time, and still lose the game. Good players change their play on each hand, sometimes run aggressive, then slow down a bit, sometimes trap, sometimes fold "Good" hands based on "trends", betting patterns, and INTUITION. Good players can recover from mistakes in math, judgment, and the others players' "lucky streak". Never will, in my judgment, a computer program have the heart, to lets say, to call an passive-aggressive player at showdown holding only a Jack high. As for the 2029 prediction, and from someone who has actually written ELIZA-like bots, from the age of 8 years old on my Commodore, to more modern "self-teaching" bots as a hobby... I've always thought that it would be possible to write a program that beats the Turing test, but never would I make the leap from a computer's math, to a living creature's intellect. If it's even possible to model and duplicate a human brain (which I believe is a fallacy), a computer will never "think" or "comprehend", it will always be a trick, a man behind the curtains, pulling the strings.

Re:Hrmmmm

[MaWeiTao]

Well, an African vulture once struck an airline at 37,000ft and other birds are known to fly at altitudes comparable to the height of Everest. Maybe not 40,000ft, but impressive nonetheless.

Re:Hrmmmm

[Anonymous Coward]

The only thing that humans do that AI doesn't (well) is automatically follow a few paths

There are a lot of differences between the brain and a CPU. To sum them up let us just say that a CPU is a computing machine while the brain is a pattern matching machine.
Sure you can compute with a brain and match pattern with a CPU, but truth is that it just doesn't work well. Some people can make three digits multiplication mentally, and some CPU with the appropriate software can recognize and categorize pictures. But he ability of a brain at doing calculus is as weak as the ability of a CPU at doing pattern matching.
A good exemple of a problem we do not know yet how to solve with computers is the following :
a) You take 50 persons and ask them to draw one or more trees using pencil and paper.
b) You take 50 other persons, completly unaware of what was asked to group a, show them the fifty drawings and ask them what is the common point between all the drawings. They will answer trees.
Now if you replace group b with a computer it won't be able to answer. If the computer doesn't have a good software or knowledge database it won't be able to answer because it won't recognize anything, if the computer has appropriate database and software it won't be able to guess that the appropriate answer is tree and not black and white drawings, paper and pencil, hand drawings and so on.
The ability that the brain has to pick up the important information among a mass of possible choices is unmatched, and the reason why it is unmatched is because the brain is capable of making arbitrary choices. In fact the brain constantly rewires itself to adapt to its environment. If you are not living in Asia, chances are that you will loose the ability to hear quater-tone notes by the age of three. On the other hand a CPU is hardwired. I never heard of a CPU dropping an internal function and reusing the freed transistors to do something else.
To put it in other terms : a brain can be wrong and learn from it. A CPU can be badly made or buggy, but it cannot be wrong. Therefore it cannot learn.

Re:Hrmmmm

[tepples]

But he ability of a brain at doing calculus is as weak as the ability of a CPU at doing pattern matching. If by "calculus" you mean derivatives and antiderivatives, solving those is pattern matching. A symbolic integral solver must recognize patterns conducive to substitution and patterns conducive to parts.

Re:Hrmmmm

[ppanon]

Passing the Turing test? (Fatally flawed because it's not double blind, btw.)

You're wrong about this, if it's properly administered. The point about a double-blind test is that the administrator of the test can't know because the subject can pick up on subliminal body language produced by the test administrator and be subconsciously affected by it, possibly triggering a placebo effect or distorted observations.

Remember that, in a Turing test, the person who's supposed to do the determination is isolated, communicating through a keyboard, and doesn't see who is on the other end. It's not like you can convince a human subject that they really are a computer simulation (although to get your AI to correctly act human, you may need to raise it in a simulated environment to make it believe it's human). So if you place a tester randomly in contact with either a human or computer subject, both types of subject will know what kind they are, but body language, voice intonations, and so on are not communicated to the person doing the avaluation. That limits it to a choice of words and, well, that's the whole point of the test isn't it?

Re:Hrmmmm

[Headrick]

African or European barn swallows? Just what are you comparing?

Re:Hrmmmm

[meta-monkey]

But can you make an airplane that that flies like a barn swallow, regardless of altitude, range or payload constraints? No, so the "airplanes have surpassed birds" bit doesn't work. The applications are completely different.

Re:Hrmmmm

[Anonymous Coward]

I'm a little less qualified than you to inform of my opinion, but I'm going to do so anyway.

Kurzweil believes in his theory entitled "the law of excelerating returns" which means that the amount of returns (milestones) in speed of processing power will rise exponentially, as it has with Moore's Law. It's one of Kurzweil's fundamental tenets and you have to believe it in order to see any value in the guy, without believing it all of this sounds rather impossible.

In the 1990s when they began to sequence the Genome, it was thought that it would take dozens of years. Because of the exponential rise in computing power, that time was drastically shortened to within a decade. I think Kurzweil would tell you that the same principle is at play here. While at our current pace of development, we will not see any major milestones, but at an exponential rate of development...

Re:Hrmmmm

[corbettw]

400 years ago...Galileo was arguing with the Pope about the Earth being flat Um, no, the argument was about the moons of Jupiter, and the fact that Galileo discovered that they orbit their primary planet, not their primary star (not entirely accurate, but a big step forward from the geo- and helio-centric universe models that came before). Considering that this discovery came over 100 years after Columbus reached the New World and several decades after Magellan* circumnavigated the globe, the question of whether the earth was flat or not had long since been put to bed.

*Actually, his crew did it, since he was supper in the Philippines.

Re:Hrmmmm

[Explodicle]

So let me get this straight... you think it would take thousands of years for a civilization with Data-like intelligences available to make an android body? I'm a mechanical engineer, and even I think the "Data's head" part is way harder.

Re:Hrmmmm

[NoOneInParticular]

As someone interested in philosophy i agree hes way off, you need to understand something before you can build it.

Before creating AI there should be some consensus as to "What is AI", "What is the mind", "What is consciousness", (and as you hint at) "What does it mean to be human" and lots of other fundamental questions.
I disagree. We're going to build increasingly complex computers for a wide array of tasks. We are going to connect them, build meta-computers to control the interconnectedness in order to define and solve sub-tasks. We're going to keep on building interfaces, and with tasks becoming more complex, the interfaces will become more complex. So we will try to create programs that create the interfaces.
At some point (way later than 2029), we will look at the whole intertangled mess and observe: "gosh, it's intelligent". At that point we might be able to define AI, not before.

Re:Hrmmmm

[bug1]

What you describe sounds line a nightmare (if i am understanding correctly)... you suggesting we are going to create so many layers on top of problems that we are going to design ways to manage all these layers and one day its going to be come so complex that we are just going to say its intelligent ?

We still need to know what intelligence means as applied to a machine, if we dont then saying it is meaningless.

Creating layers adds inefficiency and there will always be efforts by programmers to limit the indirection around solving a problem.

Until we have infinite resources i cant see your layered AI poping into existance.

I think if i saw such a mess i would proclaim a miracle rather than intelligence.

Exponential AI?

[TheGoodSteven]

If artificial intelligence ever gets to the point where it is greater than humans, won't it be capable of producing even better AI, which would in turn create even better AI, and so on? If AI does reach the level of human intelligence, and eventually surpasses it, can we expect an explosion in technology and other sciences as a result?

Re:Exponential AI?

[psykocrime]

If artificial intelligence ever gets to the point where it is greater than humans, won't it be capable of producing even better AI, which would in turn create even better AI, and so on? If AI does reach the level of human intelligence, and eventually surpasses it, can we expect an explosion in technology and other sciences as a result?

That's the popular hypothesis.

Re:Exponential AI?

[Anonymous Coward]

Go look up the Singularity. Kurzweil predicts it to be around 2040 or so.

Re:Exponential AI?

[STrinity]

If artificial intelligence ever gets to the point where it is greater than humans, won't it be capable of producing even better AI, which would in turn create even better AI, and so on?

You aren't going to get modded up for repeating things Vernor Vinge said twenty years ago.

Re:Exponential AI?

[wkitchen]

This positive feedback effect happens to a considerable extent even without machines that have superintelligence, or even what we'd usually consider intelligence at all. It's happening right now. And has been happening for as long as humans have been making tools. Every generation of technology allows us to build better tools, which in turn helps us develop more sophisticated technology. A great example from fairly recent history, and that is still ongoing, is the development of CAD/CAM/CAE tools, particularly those used for design of electronic hardware (schematic capture, PCB, HDL's, programmable logic compilers, etc.), and the parallel development of software development tools. Once computers became good enough to make usable development tools, those tools helped greatly with the creation of more sophisticated computer technology, which supported better development tools.

Superintelligence may speed this up, but the effect is quite dramatic already.

Re:Exponential AI?

[Jah-Wren Ryel]

On the other hand - perhaps human intelligence is just not at the level required to create an AI at the same, much less, better level. If that's true, we are stuck.

Re:Exponential AI?

[mdenham]

If artificial intelligence ever gets to the point where it is greater than humans, won't it be capable of producing even better AI, which would in turn create even better AI, and so on? You aren't going to get modded up for repeating things Vernor Vinge said twenty years ago. Obviously you're wrong, as he already got modded up.

Re:Exponential AI?

[wellingj]

I know there is a lot of evidence to the contrary, but what scientific law or theory says human intelligence can't advance at the same or better rate?

Re:Exponential AI?

[Excelcia]

No, there won't be an explosion, since anything we design will also inherit all our neurosis. "Brain the size of a planet, but I've got this pain in all the diodes along my left side."

Re:Exponential AI?

[Anonymous Coward]

And then Skynet becomes self-aware.

Re:Exponential AI?

[bunratty]

We already know it doesn't take any intelligence to speak of. All it takes is lots of trials, continually weeding out the bad experiments and trying new variations of the successful ones. As computers get faster and have more memory, it will take less time to try more variations, and more complex variations can be tested.

Re:Exponential AI?

[shess]

If artificial intelligence ever gets to the point where it is greater than humans, won't it be capable of producing even better AI

Depends on if it is created by understanding intelligence, or through an accident. We have one example of intelligence created by accident, which has as of yet not managed to create another intelligence.

Even understanding may not suffice. If humans create a human-plus AI and understand how they did it, then it is natural to assume that the human-plus AI could create a human-plus-plus AI, but just as there are many problems with creating a human-plus AI, there might be many problems in creating a human-plus-plus AI, and the improvements we make may not be improvements which lend themselves to the task of creating the next generation of AI. It may be that creating a super-human AI involves solving hundreds of isolated problems over a very long period, and each problem solved may only provide a tiny bit of gain in solving the next problem.

As a specific example, what if we created a human-level AI which had the additional ability to over the long run beat any human at poker? In that way, this AI would be super-human, but the AI may not provide any advantage in creation of the next generation. There are lots of specific human talents one could imagine perfecting in an AI without improving that AI's ability to create a new AI (in fact, many of those talents may be distractions from that job!).

Myself, I await the Singularity :-).

-scott

Re:Exponential AI?

[ChrisGilliard]

Yes, it is and Kurzweil predicts this will happen in 2045.

Re:Exponential AI?

[Camel Pilot]

The so called singularity which is opposed by Fermi's paradox. The greatest argument against the singularity prediction is that if true we should have detected or have been contacted by singularity-acquiring-life elsewhere in the universe.

Re:Exponential AI?

[frostband]

I think AI is already to the point where it is greater than humans.

It's just that we destroyed it before it took over the whole world.

Re:Exponential AI?

[Tablizer]

If artificial intelligence ever gets to the point where it is greater than humans, won't it be capable of producing even better AI, which would in turn create even better AI, and so on? If AI does reach the level of human intelligence, and eventually surpasses it, can we expect an explosion in technology and other sciences as a result?

This is the point when overlord jokes stop being funny.

Re:Exponential AI?

[Prof.Phreak]

Yes. Except I don't think it needs to start this trend at human intelligence.

I think we will (ie: distant future) build a very simple stupid program (nothing anyone would consider `intelligent'), and have it evolve and learn to reach and surpass human intelligence.

Sort of like humans start off with DNA, etc.

Re:Exponential AI?

[deimtee]

Ah, but you are neglecting the simulation hypothesis

Re:Exponential AI?

[aim2future]

All it takes is lots of trials, continually weeding out the bad experiments and trying new variations of the successful ones.

This sounds like evolutionary programming, even though I agree with you about experiments, this is how R&D should work, but those attempts which have been made to create AI with genetical algorithms I don't believe in. The end result may behave intelligent, but how do we know that it's intelligence is based upon ethical reasoning?

Re:Exponential AI?

[Anonymous Coward]

Uh, there is no positive feedback. It is all about reuse, levels of abstraction, structure generation and computer asisted analysis. The feedback loop of a "superintelligence" would be independent of any human involment and self-sustaining (perhaps to the bitter end).

Re:Exponential AI?

[Veilrap]

Unless you take quantum tunneling into account. Then even though our intelligence level isn't high enough to make it to the other side of the wall (ie create the ai) we might just do it anyways.

Re:Exponential AI?

[binpajama]

That is exactly what Kurzweil is predicting. See Singularitarianism

Re:Exponential AI?

[Jah-Wren Ryel]

We already know it doesn't take any intelligence to speak of. All it takes is lots of trials How do we know that? Are you seriously proposing the Infinite Monkey Theorem as being feasible? There is a reason it is called infinite you know.

Re:Exponential AI?

[Boronx]

He's proposing a variation where monkeys that make the least progress towards typing hamlet are killed off, and the remaining monkeys are allowed to breed to replenish their numbers. Conveniently, this requires only a large finite number of monkeys.

Re:Exponential AI?

[Jah-Wren Ryel]

He's proposing a variation where monkeys that make the least progress towards typing hamlet are killed off, and the remaining monkeys are allowed to breed to replenish their numbers. Conveniently, this requires only a large finite number of monkeys. No it doesn't.

Answer these questions - what's half of infinity? What's infinity divided by googol?

Re:Exponential AI?

[Boronx]

He's proposing a variation where monkeys that make the least progress towards typing hamlet are killed off, and the remaining monkeys are allowed to breed to replenish their numbers. Conveniently, this requires only a large finite number of monkeys.

No it doesn't.

Yes it does.

Answer this question: How many times have eyes evolved on this planet?

Re:Exponential AI?

[Jah-Wren Ryel]

Answer this question: How many times have eyes evolved on this planet? Once. Once in the entire infinite universe.

How many times have ESP organs evolved on this planet?
How about life that takes its fuel from atomic fusion?
Animals that move via anti-gravity?

The difference is that no AIs have ever existed regardless of origin, just as no atomic-fueled, ESP and anti-gravity using life forms have ever evolved.

So to assume that some sort of genetic algorithm can feasibly produce AI is a gigantic leap of faith.

Re:Exponential AI?

[Boronx]

Answer this question: How many times have eyes evolved on this planet?
Once. Once in the entire infinite universe.

You may be right, but many researchers think they've evolved dozens of times independently.

Fusion fuel is so far removed from the context of life that it has taken billions of years for life to learn to harness it, but it has.

Computers swim in the see of thought. It may be that algorithms are the stuff of thought. When it comes to evolving sentience, computers are playing on their home field. Is it a leap of faith to assume the right program will just evolve intelligence? Sure, but the idea is far from absurd.

Re:Exponential AI?

[Jah-Wren Ryel]

Computers swim in the see of thought. Dosing tonight?

It may be that algorithms are the stuff of thought. When it comes to evolving sentience, computers are playing on their home field. Swallowing colors of the sound I hear.

Is it a leap of faith to assume the right program will just evolve intelligence? Sure, but the idea is far from absurd. On one hand yes, on another hand no. Wave hands, wave, wave, wave. Lucy in the sky with diamonds! Yeah.

Re:Exponential AI?

[harry666t]

I think this is the way we should go. I'd even dare to say that any single human brain has much, much, much more power than all of the computers on the earth combined.

20 years is too long to predict

[Bill, Shooter of Bul]

The farther out you make a projection, the less likely it is to be true. With this one in particular, I just don't see it being a focus of research. Yes we will have increase levels of intelligence in cars toasters and ball point pens, but the intelligence will be in a supporting role to make the devices more useful to us. There isn't a need for a human like intelligence inside a computer. We have enough ones inside human bodies.

Also, I will not be ingesting nano bots to interact with my neurons, I'll be injecting them into my enemies to disrupt their thinking. Or possibly just threatening to do so to extract large sums of money from various governmental organisations.

Re:20 years is too long to predict

[Jugalator]

There isn't a need for a human like intelligence inside a computer. And even if there was (and I think this is key to the fallacy in this prediction), we wouldn't have the theories backing the hardware. We will most likely get some super fast hardware within these years, but what's much less certain is if AI theories will have advanced enough by then, and if the architecture will be naturally parallelized enough to take advantage of them. Because while we don't know much about how the human brain reasons, we do know that to make it at an as low temperature as 37 degrees Celsius in an as small area as our cranium (it's pretty damn amazing when you consider this!), it needs to be massively parallelized. And, again, we don't really even have the theories yet. We don't know how the software should best be written.

That's why we even in this day and age of 2008l, we're essentially running chatbots based on Eliza since 1966. Sure, there's been refinements and the new ones are slightly better, but not by much in a grand scheme. A sign of this problem is that they are giving their answers to your questions in a fraction of a second. That's not because they're amazingly well programmed; it's because the algorithms are still way too simple and based on theories from the sixties.

If the AI researches claiming "Oh, but we aren't there yet because we haven't got hardware nearly good enough yet", why aren't we even there halfway, with at least far more clever software than chatbots working on a reply to a single question for an hour? Sure, that would be impractical, but we don't even have the software for this that uses hard with even the boundaries of our current CPU's.

So at this point, if we'd make a leap to 2029 right now, all we'd get would be super fast Eliza's (I'm restricting my AI talk of "general AI" now, not in heuristic antispam algorithms, where the algorithms are very well understood and doesn't form a hurdle). The million dollar question here is: will we before 2029 have made breakthroughs in understanding the human brain well enough in how it reasons along with constructing the machines (biological or not as necessary) to approximate the structure and form the foundation on which the software can be built?

I mean, we can talk traditional transistor-based hardware all day and how fast it will be, but it will be near meaningless if we don't have the theories in place.

Re:20 years is too long to predict

[Jane Q. Public]

Cars and toasters are NOT "intelligent"!! Not even to a small degree. Just plain... not.

Yes, they do more things that we have pre-programmed into them. But that is a far cry from "intelligence". In reality, they are no more intelligent that an old player piano, which could do hundreds of thousands of different actions (multiple combinations of the 88 keys, plus 3 pedals), based on simple holes in paper. Well, we have managed to stuff more of those "holes" (instructions) into microchips, and so on, but but the machines themselves are just as stupid as they have EVER been, including back in the stone age. No intelligence. At all. Not even a little.

Do not mistake complexity for intelligence. A certain amount of complexity might be necessary for intelligence to exist, but on the other hand, things can be enormously complex without the presence of ANY intelligence. Just look at Government, for example.

Re:20 years is too long to predict

[Anonymous Coward]

Cars are NOT "intelligent"!!

Well then, how do they always manage to stay on the road while I'm napping!??!?

Re:20 years is too long to predict

[genericpoweruser]

A well presented and insightful comment. But consider this: what is different about a human brain? What makes it "intelligent"? That's the question that must be asked to form the theories the above commenter was talking about.

Re:20 years is too long to predict

[bitspotter]

To an industrialist, there are numerous disadvantages to humans. For one, it takes 16+ years to create a new one, you can only work them so many hours a week, they demand benefits, respect, and a come with a whole host of legal requirements to use as employees. Software is trivial to duplicate, easy to produce (and dispose of) new hardware for, is legal to abuse in ways you'd never get away with (for long) with humans, and is property.

Even dumb software workers has enormous economic value.

Now how much you want to bet working people will benefit from them?

Look, I'd love to lose my job to a machine as much as the next slacker. But how are we going to pay the bills?

The good news is that smart software can make for a huge economic boom. the bad news is that everyone working for a living now aren't likely to be included in it.

Re:20 years is too long to predict

[Degrees]

I'm going to argue that small examples of machine intelligence do exist. That is to say, there exist machines that are intelligent, however in a small space.

For example, my Subaru has a fuzzy-logic chip built into the automatic transmission. The first couple thousand miles I drove it, it was learning my driving pattern. Once it learned my driving pattern, the shifts from one gear to another were like glass.

Wall Street has A.I's that have learned to do arbitrage - see Algorithmic Trading Platforms. Read the third and fourth paragraphs to see how extensively learning A.I.'s have been deployed. This is why "day trading" is a fools market.

We have a software package at work that watches packet traffic on our network switches, and can shut down a network port if the machine attached to that port starts spewing packets at an alarming rate, or to a different set of TCP/IP ports than it "normally" does.

Each of these are examples of machine learning. Isn't the ability to learn the foundation of intelligence?

Once the foundation is conquered, the rest of it is just a matter of scale.

It's easy to discount a fuzzy-logic chip as non-intelligent, because we understand it. But I'm going to argue that just because we understand how the machine learned the patterns, doesn't mean that learning the pattern is as non-intelligent as the machines of the nineteenth century (or before). The learning is real. The goal seeking is real. The intelligence is real - it just hasn't been generalized yet.

The interesting problems: recognizing people's identity by sight/sound/patterns, comprehending language, predicting emotion - these are going to take a tremendous amount of CPU cycles. But that's just a matter of scale - not of magic. We learned it. So can the machine.

Twenty years ago, I was working on a PC that did 8 MIPS. Today, the PC I'm typing this on does 40,000 MIPS. Do I think that by 2029, the average PC will be able to learn to predict my emotion?

Yes I do.

Re:20 years is too long to predict

[Jane Q. Public]

From my own reviews of research on the subject, it appears that the main factor that is missing is a sort of "meta-awareness", or consciousness if you will, regarding information. ("Information" of course includes sensory input and surroundings). In other words, the mere possession of data is not enough; it is necessary to connect it in meaningful ways. This basic principle has been shown to correlate to the intelligence of various animals; for example, animals that have enough "self awareness" to recognize themselves in a mirror also tend to do better at other cognitive tasks. In a very real sense, they are more "conscious" than most other animals, and this appears to have a direct bearing on their "intelligence". (As an aside, it is interesting that animals that we consider to be more intelligent are also perceived by us to have more "personality".)

Re:20 years is too long to predict

[TooMuchToDo]

Cruise control =)

Re:20 years is too long to predict

[Jane Q. Public]

Quote: "Each of these are examples of machine learning. Isn't the ability to learn the foundation of intelligence?"

The simple answer is "no". The technical answer is more complex, but it still amounts to "no".

Some bingo cards can "learn", or be programmed, by sliding little windows on their faces. Flash memory "learns" in an analogous manner. A computer program can be made to "learn" about future conditions... and respond to those conditions. HOWEVER, its "learning consists only of storing information about its environment. The computer acts on this stored information in completely predictable ways... predictable because they were PRE-PROGRAMMED into the machine.

Exactly the same is true of the stock-trading programs. All they do is manipulate data that is fed to them, in completely predictable ways.

Honeybees behave and interact in extremely complex ways, and even learn about their environment. Their behavior is much more complex and adaptive than that of any modern automobile. But they are NOT intelligent... they operate by instinct ("pre-programmed behavior"). One might argue about where that programming originally came from, but that is not relevant to the discussion at hand. They are instinctive: pre-programmed.

In that sense, the automobile -- and stock market software -- display entirely "instinctive" behavior. To a certain degree, they can alter behavior depending on the conditions, but they do not "decide" to do so... they were told to do so by some progammer, much like me, sitting behind a desk somewhere.

Re:20 years is too long to predict

[Anonymous Coward]

And who will buy the industrialists products? The economy is not as simple as that. If AI's took away all the "workers" jobs then they would also be out of a job very quickly.

Re:20 years is too long to predict

[tsjaikdus]

>> Also, I will not be ingesting nano bots to interact with my neurons, I'll
>> be injecting them into my enemies to disrupt their thinking.

If, unlike you, your enemies are willing to ingest nanobots, I wish you the best of luck with your quest.

Re:20 years is too long to predict

[tsjaikdus]

>> I'm going to argue that small examples of machine intelligence do
>> exist. That is to say, there exist machines that are intelligent,
>> however in a small space.

I think those examples are not interesting. Dolphins are intelligent too, though hard to communicate with. Extra terrestrial bacteria are interesting only if that increases the chance of finding technical advanced civilizations. We are looking for others like us.

Re:20 years is too long to predict

[greyhueofdoubt]

>>There isn't a need for a human like intelligence inside a computer.

I beg to differ. A video game that allowed me to hunt genuinely intelligent human targets, a la 'the most dangerous game', would be the killer app.

It could be a bloody-fanged, spine chilling, otherworldy experience, or it could be soul-crushing, depending on how well you did. Getting your ass kicked in UT2026, where the opponents are real, and smarter than you, could suck for your self-esteem. Or how about a game in which you race the AI Risk- style?

We have online games now with real humans, but this AI would not be a 14-year old. It would not think that it was in a game. It would fight to the death in new and clever ways. It would get to know you. And I don't think it's morbid to say these things; look at the state of gaming now.

If anything, it would be good training for law enforcement/soldiers//. mods...

-b

Re:20 years is too long to predict

[tfiedler]

Really? I would think the opposite. I'm pretty sure that in, say 1000 years, humans will either be extinct (my bet is on this one) or living amongst the stars on other planets. I may be a romantic but I'd put money on the second happening. Now, if someone said that in 30 years we'll be living on Mars, I'd laugh in their face.

Re:20 years is too long to predict

[Degrees]

I guess you and I are going to have to agree to disagree.

I think I can see where you are coming from. Is this accurate? Your opinion is that if the goal-seeking behavior is supplied by an external mechanism, then the object is not intelligent, but rather just executing instinctual patterns.

In other words, innate behavior is not intelligent, only learned behavior is intelligent.

From my point of view, if it looks like a duck, and acts like a duck, and swims like a duck, and can learn the things a duck learns, it's probably a duck - or close enough that the differences don't matter.

I do expect that we will soon have software that will auto-optimize itself, adding new goal-seeking behavior of it's own discovery. But I admit - I don't know of any real-world examples today.

Re:20 years is too long to predict

[Degrees]

I think they are interesting in the same way that the first iron smelters are interesting. Sure, what came out of them was really crude - but they did boot-strap a significant advancement in technology. All subsequent ages were more advanced, because refined metal was now a possibility.

Re:20 years is too long to predict

[Jane Q. Public]

Well, no, that is not what I am saying, but what I am saying is not easy to communicate.

I am not trying to distinguish between instinctive and learned behavior. I am trying to distinguish between instinctive and "intelligent" behavior. That is exactly one of the points I was trying to make: a car's computer might "learn", but that does not make it intelligent.

Not all -- or perhaps even most -- intelligent behavior is learned. The more intelligent the creature, the more variable its behavior becomes, precisely because it is NOT proceeding along strict pre-programmed lines. It can decide what to do. And I mean "decide" in the true sense -- to make a conscious choice -- as opposed to following pre-programmed rules about how to behave given certain inputs. In other words, it has, to a certain degree, "free will". It can decide to act AGAINST any or all of the behavior it has previously learned, if it wants.

You cannot say that about a computer. Even if I put a random-number generator in there, so that it "chooses" differently sometimes, I am still controlling its behavior. I have told it, in effect, to "do this other thing 37% of the time" or whatever. It is still not free will, and it is still firmly programmed.

The distinction might be hard to describe, but to me it is very clear.

Re:20 years is too long to predict

[bythescruff]

Twenty-three billion neurons ought to be enough for anybody.

Re:20 years is too long to predict

[Degrees]

Fair enough. I see where you are coming from, with the Free Will thing.

We'll have to take up the discussion again when I can simulate that. ;-)

Some major assumptions

[Yartrebo]

How are we so sure that advances in computers will continue at such a rapid pace. Computer miniaturization is hitting against fundamental quantum-mechanical limits and it's crazy to expect 2008-2028 to have progress quit as rapid as 1988-2008.

Short of major breakthroughs on the software end, I don't expect AI to be able to pass a generalized Turing Test anytime soon, and I'm pretty certain the hardware end isn't going to advance enough to brute-force our way through.

Re:Some major assumptions

[tiffany98121]

because quantum computing is going to blow away our current rate of increase in computing power.

Re:Some major assumptions

[_KiTA_]

How are we so sure that advances in computers will continue at such a rapid pace. Computer miniaturization is hitting against fundamental quantum-mechanical limits and it's crazy to expect 2008-2028 to have progress quit as rapid as 1988-2008.

Short of major breakthroughs on the software end, I don't expect AI to be able to pass a generalized Turing Test anytime soon, and I'm pretty certain the hardware end isn't going to advance enough to brute-force our way through. Stuff already passes Turing Tests.

Sort of, anyway.

Re:Some major assumptions

[Anonymous Coward]

How are we so sure that advances in computers will continue at such a rapid pace. Computer miniaturization is hitting against fundamental quantum-mechanical limits and it's crazy to expect 2008-2028 to have progress quit as rapid as 1988-2008.

And no modem can ever exceed [110,300,2400,9600,14.4k,28.8k,33.6k,56k] (delete as appropriate for current year) baud on a voice line, right?

Re:Some major assumptions

[protohiro1]

Very interesting, because it looks like 56k was basically the hard limit. Unless you have some new, superfast modem I haven't heard of.

Re:Some major assumptions

[olman]

Nah. For example there are no ways to do proper 3-dimensional cores yet. I should find out but I think you cannot put transistors on top of each other with the current manufacturing technologies.

And a surprisingly big area of that chip space is in fact cache ram. "Empty" area on the left is 4MB cache. Nobody asked to standardize on a 4-register design for petes sake. Yes, I know, modern 64bit processors more or less emulate the x86 assembler and do the things completely differently internally. But I daresay the PPC 64-register architechture would do fine with an order of magnitude less cache ram.

You're really limited by heat and frequency and actually making things smaller helps on both. But still, note how the trend has gone away from megahurz wars to number of cores. I'd hazard that in the long run that's actually a good thing and should make CPU usage more energy efficient etc because you have most of the cores idling when not needed.. Now if we just get the damn games written to be multicore-friendly!

To heck with Artificial Intelligence!

[RyanFenton]

Artificial intelligence would be a nice tool to use to reach towards, or to use to understand ourselves... but rare is there a circumstance that demands, or is worth the risks involved with making a truly intelligent agent.

The real implication to me, is that it will be possible to have machines capable of running the same 'software' that runs in our own minds. To be able to 'back up' people's states and memories, and all the implications behind that.

Artificial intelligence is a nice goal to reach for - but it is nothing compared the the siren's call of memories being able to survive the traditional end of existence, cellular death.

Ryan Fenton

Re:To heck with Artificial Intelligence!

[clusterlizard]

It would come in pretty handy for space exploration.

Re:To heck with Artificial Intelligence!

[Fred Ferrigno]

The real implication to me, is that it will be possible to have machines capable of running the same 'software' that runs in our own minds. To be able to 'back up' people's states and memories, and all the implications behind that. Maybe some famous person would agree to do that as a favor to future generations, but I don't see how it would really be useful for average people. We are not the "software" that our brains run, we are our brains. The copy of "you" that runs on a computer for thousands of years won't really be you. You'll still be dead.

Re:To heck with Artificial Intelligence!

[Bob3141592]

The real implication to me, is that it will be possible to have machines capable of running the same 'software' that runs in our own minds. To be able to 'back up' people's states and memories, and all the implications behind that. That presumes you can understand how human thought is made. It presumes real human intelligence can be modeled and implemented by a digital process, which may not be possible. I doubt that even quantum digital computers could do it. It might be possible in the future to simulate our neural machinery without realy knowing how it works, a high-fidelity digital form of a completely analog process, but then you couldn't know what to expect as the result. The way the program was coded and the inputs given it would have no predicitive value. After all, as far as we know a fool's brain is made up just like a genuis's. Of course, I'd be impressed if they could artificially recreate even very foolish human intelligence. I fear that may be the only kind we will ever do. remember, nothing is more frightening than ignorance in action.

Re:To heck with Artificial Intelligence!

[ricree]

I'd rather a copy live on than nothing at all.

Re:To heck with Artificial Intelligence!

[monoqlith]

I've given this a fair amount of thought, and am frankly suspicious, nay, fairly certain, that uploading your mind to a machine isi mpossible. Actually, not impossible - inconceivable, one step above impossible.

Sure, it may be possible to upload the information "contained" in your mind - your declarative and procedural memories and so forth - to a machine to be used later in some practical fashion. However, I'm of the persuasion that individual consciousness is very much tied to the organizational structure of a particular set of atoms - namely, the atoms that compose our brain, and that thus there would be no way to "upload our mind" to a machine such that we would then be able to continue our life and experience things as ourselves beyond the death of our body.

I don't think "software" so much as runs on the brain as it is *hardwired* into the way the brain works - the computer/hardware/software model is not IMHO useful way of thinking about the brain. There are philosophers who would disagree with me, of course, and they have gained a certain amount of traction. I am not a neuroscientist, or a philosopher, just a lowly philosophy major writing a thesis on philosophy of mind, but I do have a brain disease, and this has given me first-hand insight into way our mental states relate to our physical brain states.

The brain is designed from the ground up to be an analog computer of sorts. It processes things the way it does because it is physically organized to do so, not because software was "written" for it that it somehow serves as the central processor for.

This is not to say that we can't create a sentient, intelligent machine using hardware/software. I believe that we can. At the basis of the brain there is still an algorithm that we can instantiate in software. But the consciousness we create would be a unique one, it might not be guaranteed to think the way we think, and there would be no way to verify that its "qualia"(qualitative experience) would 1) exist or if it does exist 2) be like ours in any way.

Re:To heck with Artificial Intelligence!

[DamnStupidElf]

The copy of "you" that runs on a computer for thousands of years won't really be you. You'll still be dead.

Just as dead as the meat copy of you from 5 minutes ago? What magic makes your body 5 minutes in the future "you" instead of just a random copy? You do know that all your atoms get replaced every few years, and that when you sleep deeply or get put under general anesthesia almost all of your brain activity ceases, right? I have no problem with going to sleep at night and waking up in a slightly different biological body that thinks it's the same person as the one that went to sleep last night. Why should I care if the body I wake up in is made out of electronics instead of meat, so long as it feels the same way?

Re:To heck with Artificial Intelligence!

[RyanFenton]

I don't disagree that what you argue here may be true... but I would argue that is also might not make a difference.

First off, presuming some detailed information can be read from the mind, let's say that you can never replicate a person's memories and brain together to the point of having that same person even casually in reality again. Fine - then you still would presumably have the ability to store the data and query it. That means, you don't get Grandma back, but you can still 'ask grandma a question', which still makes for a profound change to society, and the countless things we don't notice we lose with each individual.

Next, let's say that a simulacrum of an individual's mind and personality can be replicated in some kind of computer-like environment. Even if it's not the same person, it is the same personality, and the same memories. Society may not have the same person - but they have something that can last longer than a mere 100 years, that can carry real human experience further than ever before.

Next, let's say that a mind can be restored through copying to a brain - but it's just a copy-person. I don't see how society couldn't quickly adjust to considering such restored copies as the same as the original that was lost. Just like the Star Trek crew after beaming-up back to the ship - an atomically new captain, but in practice is the same trusted mind as ever before, even if the ostensibly 'real' captain was disintegrated long ago.

What ends up actually panning out should be interesting. But I think the draw of developing the ability to copy minds, even if fruitless, is far more inherently compelling than even the allure of artificially intelligent agents - though you may end up needing one to develop the other.

Ryan Fenton

Re:To heck with Artificial Intelligence!

[Fred Ferrigno]

I don't know how the continuity of consciousness works to say that I am the same consciousness from 5 minutes ago, 5 days ago, or 5 years ago. Then again, neither do you know how it works to say that I'm not the same consciousness. One thing is clear though: the "software" copy would definitely not be the same consciousness. It's not that you will wake up in an electronic body, it's that you won't wake up at all, because you are your meat body. Even if that means you are only your meat body for the next day or few years, you will never be the electronic body.

Re:To heck with Artificial Intelligence!

[sacrilicious]

Why should I care if the body I wake up in is made out of electronics instead of meat, so long as it feels the same way?

Because it might be harder to get laid?

Re:To heck with Artificial Intelligence!

[jdoeii]

> I don't know ... [o]ne thing is clear though: the "software" copy
> would definitely not be the same consciousness

You don't know but it is clear :-). Sure.

Suppose one neuron in your brain is replaced by a software emulation. Are you the same now or different? If you say different, please explain, because lots of neurons die during your life time. If two neurons are replaced? Three? Million? Billion? Where is the line?

See, not so clear anymore :-)

Re:To heck with Artificial Intelligence!

[smannell]

If all of my atoms get replaced every few years, how come I have a tattoo that is over 15 years old and looks nearly identical to the day I got it? I don't think my body is replacing the ink. I realize the atoms in the ink were added to my body, but so were the ones in my bones and teeth. I also have scars and moles that have been the same for decades. I'm not saying that cells aren't constantly multiplying and dying, but I've always wondered if a lot of the atoms get recycled locally, and therefore some have been there for a long time.

I agree that it doesn't matter if my body changes overnight or my brain shuts down during surgery, I'm still me. A perfect copy of me wouldn't know it was a copy, and wouldn't care.

I'm just curious to know how often the body replaces atoms; especially metals and in places like tooth enamel, bones, nerves, etc.

Re:To heck with Artificial Intelligence!

[Fred Ferrigno]

If someone came to me with two dogs and asked me to determine if they were the same breed, I might have some trouble answering them. If they came to me instead with a mutt and a park bench, the answer would be clear. No one knows for sure where the dividing line is exactly, but there is at least an upper bound: without any part of your previous brain, your consciousness can't possibly persist.

Also, I would say that 'software' emulation in the case you describe is a misnomer. It may be that artificial neurons could slowly replace biological ones and retain the sense of consciousness. (It may also be that your consciousness actually does die a little with each neuron, replaced slowly by a new and different consciousness.) However, the software powering the neurons would not be where consciousness lies. As an abstract concept, software cannot "do" anything. It takes hardware to execute the software and actually perform the action. If we should ever figure out how to create consciousness through software, then it would still be the hardware that experiences consciousness.

The point is that we are not software, but hardware. If you replace that hardware outright, as the OP suggested, you will certainly cease to exist.

Re:To heck with Artificial Intelligence!

[clusterlizard]

Let's say at some point in the future we have the ability to repair damage to the brain with computing devices. Say I have a stroke and it impairs my vision, I go in for a procedure that replaces the damaged part of my brain with a device that mimics that functionality, restoring my vision. Am I still me? I have another stroke and lose control of the right side of my body. Go in for another procedure to fix that. Am I still me? As I age, this goes on (I eat at McDonald's four times a day). At what point am I gone and turned into just a copy of who I used to be running on different hardware? Our cells are being replaced constantly. I don't think it's as cut-and-dried as you make it out. On the other hand, uploading my mind in bulk into a system and expecting that to be "me" doesn't intuitively seem possible. A consciousness can't exist in two different places... or can it? Maybe it seems impossible only because we haven't experienced it. Whatever the case, it's a fascinating subject.

Re:To heck with Artificial Intelligence!

[DamnStupidElf]

Also, I would say that 'software' emulation in the case you describe is a misnomer. It may be that artificial neurons could slowly replace biological ones and retain the sense of consciousness. (It may also be that your consciousness actually does die a little with each neuron, replaced slowly by a new and different consciousness.) However, the software powering the neurons would not be where consciousness lies. As an abstract concept, software cannot "do" anything. It takes hardware to execute the software and actually perform the action. If we should ever figure out how to create consciousness through software, then it would still be the hardware that experiences consciousness.

First, changes in consciousness happen all the time; it's called growth and learning. The child that you think you were several years ago no longer exists in any meaningful sense because that consciousness grew and changed into the present you. Every second our consciousness changes and leaves the old behind. Quantum mechanics ensures that it's actually impossible to retrace the exact steps that lead from an earlier state of consciousness.

Second, your distinction between hardware and software is meaningless in this context. When a computer runs software, only the hardware exists in the physical sense. So how does the software exist? It's just a portion of the configuration of the hardware, e.g. an aspect of the hardware state. The same is true of the brain and of a brain simulation. Whatever hardware runs a brain exists physically, and the software is encoded in the physical relationships of the atoms making up the hardware. I guess what I'm trying to say is that only hardware exists, but software is the word we use to talk about the important relationships inherent in the hardware.

The point is that we are not software, but hardware. If you replace that hardware outright, as the OP suggested, you will certainly cease to exist.

What does it mean to cease to exist? The relationships between your atoms are irreversibly changed? If so, that happens every instant. Is it merely the degree of change? Would every atom of your body have to settle to maximum entropy, or would a smaller change suffice? What changes do not cause a person to cease to exist? If the change from meat to electronics causes less of a change than the level that would qualify for cessation of existence in a meat body, wouldn't a person continue to exist as electronics?

Existence is not a trivial concept since it is a completely arbitrary human definition lacking exact specification. If you can define the reason that your body, which is different every second, does not cease to exist, then I can define a transformation to an electronic body that meets all the properties you list. It's just a matter of how sensitive and accurate the hardware simulation needs to be.

Re:To heck with Artificial Intelligence!

[Anonymous Coward]

There was an interesting special on the National Geographic channel a few months ago about the growth and development of the human body. I can't remember the exact name of it though. The total replacement of all cells in the body was stated to be something like every 15 to 20 years, with total replacement slowing down after puberty and continually decreasing until complete cessation-- death.

Scars are a result of humans having the inability to regenerate tissue. While it is true that humans do have the ability to regenerate in some sense of the term, such regeneration requires a dieing cell to duplicate before its preprogrammed death. If it can't, in the case of an injury, the body creates somewhat non-specific scar tissue which is inferior in quality. In early evolutionary times our inability to regenerate in instances of injury was actually an evolutionary advantage due to the faster healing times it allowed, since it doesn't require determing which specific cells needed to be regrown-- a patch job really. Moles are the same in their presence although their reason for emergence is different.

I'm not entirely sure why tattoos remain. I'd assume that since cellular replacement is so gradual that the ink between the cells is largely unaffected to the point that it is nearly or entirely undetectable by the eye.

Re:To heck with Artificial Intelligence!

[monoqlith]

Notice how I said "organizational structure" - not the atoms themselves. What you say about cells being replaced is true - but even more often than entire cells are replaced, the atoms that compose our cells are being cycled in and out constantly. The set of atoms that make up our bodies - brain and all - today is not the same set as the one that made it up when we were born or even when we were a couple years younger. So I am by no means attaching the individual consciousness to a particular set of atoms - rather I believe that it is the organizational structure(the physical relations) of those atoms over time that allows for the emergent phenemonon we refer to as "consciousness" to happen. The organization of our brain gives us the clear idea that we persist, we feel, we experience from one moment to the next.

Re:To heck with Artificial Intelligence!

[monoqlith]

I understand now that you may try to point out my contradiction. When I said "particular set of atoms" the first time I mean - the set of atoms X that comprises our brain at time Y. This set may not be the same set that comprises our brain at time Z. However, as long as the substituted particles maintain similar relations to the overall structure of our brain, the continuity of our experience persists.

Re:To heck with Artificial Intelligence!

[Raenex]

Why should I care if the body I wake up in is made out of electronics instead of meat, so long as it feels the same way? But what does it mean when two "copies" of you exist at the same time, each having their own experiences? Is that one conscience or two? Which one did you "wake up" to?

Re:To heck with Artificial Intelligence!

[DamnStupidElf]

But what does it mean when two "copies" of you exist at the same time, each having their own experiences? Is that one conscience or two? Which one did you "wake up" to?

I'd wake up in both bodies as two separate consciousnesses with the same memories. If the copies never met each other, they wouldn't know the difference. "I" would almost certainly diverge in experiences in each copy just from random quantum effects, not to mention environmental differences, until "I" was no longer classifiable as a single individual. It may take seconds, or it may take weeks, depending on how similar our environments were. If we started out in a perfectly symmetrical room facing each other (and the universe was deterministic), it could even be argued that we would remain the same person since there would be no way to introduce differences into our brains. We'd react to each other, but if the timing and motions were perfectly exact we'd just keep doing the exact same things at the same time.

Actually, if you believe the Copenhagen interpretation this happens all the time in alternate universes spawned by the collapse of quantum states, except that there are infinitely many copies going their own way every instant.

Re:To heck with Artificial Intelligence!

[Raenex]

So, if you wanted to get from Los Angeles to New York, would you step into the booth?

Re:To heck with Artificial Intelligence!

[Fred Ferrigno]

The child that you think you were several years ago no longer exists in any meaningful sense because that consciousness grew and changed into the present you. You state this as fact with no way to back it up. The truth of the matter is neither of us can define "consciousness" exactly to say that I do or do not share it with a previous version of myself. However, an exact definition is not necessary to say concretely that a completely separate entity would not share that consciousness. The proof is easy enough: there is nothing about making a software copy that would preclude your meat body from continuing as a separate entity.

It's just a portion of the configuration of the hardware, e.g. an aspect of the hardware state. ... I guess what I'm trying to say is that only hardware exists, but software is the word we use to talk about the important relationships inherent in the hardware. And what I'm trying to say is that simply copying the configuration of that hardware to completely different hardware cannot possibly transfer consciousness.

What does it mean to cease to exist? Again you try to laden me with an impossible burden and again I say an exact definition is not necessary. It suffices to say that the end of consciousness in this sense is at least the same as with natural death, whatever that may be, because the process is identical. Copying the software of a person's brain would have no affect on their subjective experience of life. In fact, it might even be done without their knowledge. They would simply live the rest of their life for however long it lasts, then eventually die as they would anyway. That a separate entity exists with their personality and memories is irrelevant to that process.

If you want to say that consciousness ceases to exist before a person dies, then the burden is on you to come up with a different, more precise definition and substantiate that with evidence. Perhaps it does. I don't know, but then neither do you.

Probably false alarm ... again

[golodh]

For over 40 years, the field of AI has been *littered* with predictions of the type: "We will be able to mimic Human levels of xxx" (substitute for XXX any of the following: contextual understanding, reasoning, speech, vision, non-clumsy motoric ability).

So far _not one_ of those claims has come true, with the possible exception of the the much-vaunted "robotic snake".

So ... I'd say: less claims, fewer predictions, and more work. Let me know when you've got anything worthwhile to show.

Not to be outdone by forecasters, I have a forecast of my own to make: before the term is us it will transpire out that all this fanfare and this announcement were only ever meant as means to attract research grants.

Re:Probably false alarm ... again

[lambent]

I concur, except replace 'probably' with 'definitely'.

We have not accomplished any real advances that bear even a faint similarity to what this prediction purports. This is just passing the buck, and hoping that in the future, someone will solve the problem we have not even made any real progress against, yet.

In the end, this is just idiot blathering. It doesn't help further any effort, and is just so many useless words.

I'll make a prediction, too. By 2029, we will realize we've wasted another 20 years trying to achieve human level AI in computers, but people will continue to make wild claims about what will happen 20 years out from that.

In 2029, they'll say, "by 2049, for sure. This time we really really mean it."

Re:Probably false alarm ... again

[TrevorB]

Agreed. Turing test passing AI is way up there beyond flying cars in predictions that won't come true for a very, VERY long time, if ever. If we're going to see an AI in 20 years, it's going to be biological rather than electronic. This is a problem where no amount of extra computing power you can throw at the problem will solve it. The theory to make it work just isn't there.

If you could show that a machine (say like SETI at home) could pass a Turing test, even if it needed a day to make a reply, I'd believe it was possible. It's just not coming, folks.

Re:Probably false alarm ... again

[timeOday]

So ... I'd say: less claims, fewer predictions, and more work. Let me know when you've got anything worthwhile to show.

Has it occurred to you that all of us already work, to some extent, at the direction of computers? Think of the tens of thousands of pilots and flight attendants... what city they sleep in, and who they work with, is dictated by a computer which makes computations which cannot fit inside the human mind. An airline could not long survive without automated scheduling.

Next consider the stock market. Many trades are now automated, meaning, computers are deciding which companies have how much money. That ultimately influences where you live and work, and the management culture of the company you work for.

We are already living well above the standard that could be maintained without computers to make decisions for us. Of course as humans we will always take the credit and say the machines are "just" doing what we told them, but the fact is we could not could not carry out these computations manually in time for them to be useful.

Re:Probably false alarm ... again

[Anonymous Coward]

That's what they said about flight in 1902.

Seriously, there were definitely grandiose promises made in the early days of AI, before anybody had any idea how hard it would be, but there has been incredible progress in those 40 years. Until you succeed, it always looks like you've failed and will continue to fail.

Re:Probably false alarm ... again

[Alomex]

In other words be lear of any field in which the prefered mode of publication is the press release e.g. AI a la MIT Media Lab, complexity (Santa Fe Institute), semantic web, etc.

Re:Probably false alarm ... again

[Anonymous Coward]

Of course as humans we will always take the credit and say the machines are "just" doing what we told them, but the fact is we could not could not carry out these computations manually in time for them to be useful.

And thus *they* are doing it of their own volition, or *they* should receive the credit? We could say the same thing of any machine -- a lawnmower, for instance -- we just couldn't carry out the task of mowing a golf-course in the time it would take for it to be useful, therefore lawnmowers should receive the credit, not humans?

2029? Skynet? Coincidence?

[Rich Acosta]

From Wikipeida

Skynet gained access to several autonomous military drones (such as the T-1 in Terminator 3), using them to round up survivors, who were forced to build automatic factories and robots that were better at construction than the military robots. Skynet then killed these human slaves, and using the infrastructure they had been forced to start, rapidly designed newer and better machines until it controlled an extremely advanced empire on Earth by 2029.

This is good news and bad news...

[zappepcs]

Good news: This could herald a lot of good stuff, increased unemployment, greater reliance on computers, newer divides in the class strata of society, further confusion on what authority is and who controls it, as well as greater largess in the well meaning 'we are here to help' phrase department.

Bad news: After reviewing the latest in the US political scene, getting machines smarter than humans isn't going to take so much as we thought. My toaster almost qualifies now. 'You have to be smarter than the door' insults are no longer funny. Geeks will no longer be lonely. Women will have an entire new group of things to compete with. If you think math is hard now, wait till your microwave tells you that you paid too much for groceries or that you really aren't saving money in a 2 for 1 sale of things you don't need. Married men will now be third smartest things in their own homes, but will never need a doctor (bad news for doctors) since when a man opens his mouth at home to say anything there will now be a wife AND a toaster to tell him what is wrong with him.

oh god, this list goes on and on.

Re:This is good news and bad news...

[Anonymous Coward]

'You have to be smarter than the door' insults are no longer funny.

You watch this door, it's about to open again. I can tell by the intolerable air of smugness it suddenly generates.

I'm not getting you down am I?

Re:This is good news and bad news...

[hawkfish]

After reviewing the latest in the US political scene, getting machines smarter than humans isn't going to take so much as we thought. My toaster almost qualifies now.

"The strongest argument against democracy is a five minute conversation with the average voter." - Winston Churchill

Re:This is good news and bad news...

[Anonymous Coward]

If you think math is hard now, wait till your microwave tells you that you paid too much for groceries or that you really aren't saving money in a 2 for 1 sale of things you don't need.

if my microwave gives me any shit, I'll demote it to tampon dispenser

Re:This is good news and bad news...

[zappepcs]

You insensitive clod - quote from my wife's tampon dispenser

AI may not get that far

[httpcolonslashslash]

As soon as they make robots that can have sex like humans...what's the point in inventing anything else? All scientists will be busy "researching" their robots.

Re:AI may not get that far

[weighn]

Artificial Insertion?
Piers Anthony had some interesting ideas on this ...

Re:AI may not get that far

[imasu]

Or as Scott Adams put it, "[The holodeck] will be society's last invention."

Re:AI may not get that far

[Lord Ender]

This is why we need to start a religion that abhors interactive entertainment (video games today, holodeck eventually). Something like the Amish.

Once our population dwindles as we game away our lives, these religious nuts can repopulate the earth.

Re:AI may not get that far

[Your.Master]

Why should you care if the earth is populated, if you and those you love are already gone with no children?

Re:AI may not get that far

[Lord Ender]

Who cares about the future? You must be a Republican.

2029?

[olrik666]

Just in time for AI to help me drive my new fusion-powered flying car!

O.

Re:2029?

[IdeaMan]

Just in time for AI to help me drive my new fusion-powered flying car! THIS is what you get for not washing me! (Lowers windows while raining.)
THIS is what you get for looking at my best friend Amber Chevy like that. (Retracts roof and unfastens seatbelt).
And THIS is what you get for taking that long ride with Claire Maserati!! (Flips over).

Not sure that I want my car being flown by AI...

wrong

[j0nb0y]

He obviously hasn't been paying attention to AI developments. The story of AI is largely a story of failure. There have been many dead ends and unfulfilled predictions. This will be another inaccurate prediction.

Computers can't even defeat humans at go, and go is a closed system. We are not twenty years away from a human level of machine intelligence. We may not even be *200 years* away from a human level of machine intelligence. The technology just isn't here yet. It's not even on the horizon. It's nonexistent.

We may break through the barrier someday, and I certainly believe the research is worthwhile, for what we have learned. Right now, however, computers are good in some areas and humans are good in others. We should spend more research dollars trying to find ways for humans and computers to efficiently work together.

Re:wrong

[Anonymous Coward]

He's not talking about the software shortcuts that most of us are used to regarding AI we see in real life. He's talking about simulating every aspect of the human brain down to every neuron in software. With the current exponential trends in computing that seemingly keep breaking every barrier that people keep referring to, it only makes sense to extrapolate that out to see what the logical conclusion is. They've already started simulating portions of rat brains in software that behave identically to the real counterpart.

Re:wrong

[Alaria Phrozen]

Excuse me.. who are you? You're saying RAY KURZWEIL hasn't been paying attention to AI developments? And you're modded insightful?

http://en.wikipedia.org/wiki/Ray_kurzweil

"Everybody promises that AI will hit super-human intelligence at 20XX and it hasn't happened yet! It never will!" ... well guess what? It'll be the last invention anybody ever has to make. Great organizations like the Singularity Institute http://en.wikipedia.org/wiki/Singularity_Institute really shouldn't be scraping along on such poor budgets, seriously if this ever worked, even a 0.001% chance of a friendly technological singularity occuring, isn't it worth investigating?

Re:wrong

[Anonymous Coward]

According to wikipedia, Ray Kurzweil is 60.

It's obvious that he is just desperate for the singularity to start before he dies. It is wish fulfillment at its worst, leading to poor predictions.

As he has aged, he has unfortunately become wackier and wackier in his attempts to stave off death*, and his wish for technological AI advances are part of this (as he considers AI advancements necessary for a technological singularity which will give people effective immortality).

*(Heck, I read on wikipedia that he touts something called alkaline water for health benefits. I'm a biochemist with my PhD in oxidative radical chemistry and ageing, and it is obvious that he understands almost nothing of biochemistry and chemistry from the linked article. It's sad when smart people go off the deep end, but it happens to a surprising amount of them).

Re:wrong

[Cedric Tsui]

Hmmm. Interesting.

But when can it be said that a computer is smarter than a human? A computer can do certain things better than a human can, but irregardless of how many transistors you have, they will never be able to do everything better than a human.
Just what will a computer need to be able to do before we say; yup. They've surpassed us.

I think 200 years is a pretty far cry, and would say that 20 is closer. Computers have only been around for about 200 years (they had punch card controlled looms 200 years ago didn't they?)

Think about the high performance computing labs these days. I can't predict global weather patterns. Can you?

Re:wrong

[spintriae]

According to that Wikipedia article, he previously predicted that a $1000 computer would be 1000 times more powerful than the human brain by 2029. Now he's simple stating there will exist a computer of human level by 2029. Hopefully by 2029 there will exist a computer powerful enough to correct these phony calculations.

Re:wrong

[giorgist]

;-) You know the anthropomorphizing is your flaw. It may take 2000 years to beat humans in GO but by that measure computers can beet humans in a myriad of tasks. Humans will never be able to compete. The list is growing bigger all the time. Computers or automated machines can do better in agriculture, transport, recoding of knowledge, retrieving knowledge and on an on ... Simply put you placed the marker at the Human level but you forget that Google can be thought of a sentient being. It employees people as benevolent parasites and just grows. It does not have sense of self, but you assume humans do. Computers will become as intelligent as humans but for a very short period in time. After that it will have access to all of human knowledge, all of human creation. It is either scary or liberating G

Re:wrong

[glwtta]

Excuse me.. who are you? You're saying RAY KURZWEIL hasn't been paying attention to AI developments?

I don't need Wikipedia for this, just from the linked article I can tell that Ray Kurzweil either (a) has no clue about current AI developments, or (b) willfully misrepresents them in order to sell his crappy science fiction posing as "futurism".

Which one is it?

Re:wrong

[glwtta]

you forget that Google can be thought of a sentient being

Well, sure. In the same way that I can be thought of as dating Heidi Klum.

No one can actually force you to restrict your thoughts to things that have some sort of grounding in reality.

It does not have sense of self, but you assume humans do.

Heh, isn't sense of self the one and only thing that philosophy has manage to actually prove over the last 2,500 years?

Re:wrong

[Anonymous Coward]

The story of flight was also a story of failure. Until it wasn't.

Re:wrong

[jberryman]

It'll be the last invention anybody ever has to make. Odds are that we are that we are all right now just computer-simulated AI entities in a vast "ancient ancestor simulation" created by the lifeforms that we believe we will become in several hundred million years. In fact we may be simulations created by simulations created by simulations. If this seems unlikely, consider that any creature with the capability of creating such a simulation logically would do so in order to learn about it's past, maybe even for fun (imagine another parallel universe in some future-hacker's basement).

Assuming that such a simulation is not a logical impossibility, and that there is not an insurmountable barrier to intelligence, (and a few other things) it becomes far more likely that we don't even exist.

Re:wrong

[glwtta]

But when can it be said that a computer is smarter than a human?

When a computer can do something better than a human without first being programmed by a human with a way to do it, and then operated by a human to actually perform the task?

Let's take everyone's favorite example: chess. Here's how I would define the point where computers surpass humans at chess:

1) A team of engineers is given the task of designing an AI. No one says a word about any specific applications for it.
1a) Once the programming is complete, the AI has no further interaction with its designers, of any kind.
2) The AI is given the rules to chess. In English.
3) The AI can reliably defeat human opponents of moderate ability.
4) After some amount of practice, and possibly hardware expansion (but not chess-specific hardware), a single instance of such an AI can reliably defeat the reigning human world champion.

Does that sound like a reasonable time to concede that AI has surpassed humans in chess?

How many decades away do you think we from this happening?

Re:wrong

[Tablizer]

Computers can't even defeat humans at go

They used to say the same about chess.

Re:wrong

[castrox]

The current AI involved with Chess (and Go) are search-based. To me, search-based AI is just a dumb AI which can at most say that one strategy is better than another - which mostly is based on heuristics.

Afaik Big Blue had no reasoning behind its moves - just a vast bank of possible moves of which some were rated as better based on previous games by other master players.

It's even theorized that a semi-pro could have been more successful meeting Big Blue since the game style Big Blue used was on par with Kasparov which is different from a semi-pro.

Also, go is a number of magnitudes worse in complexity than chess (the larger spaces) which is why it is tough to "solve" - we can't store the entire solution set or search through it fast enough.

Search-based AI = stupid AI. Albeit efficient and sufficient at many tasks, it's not real AI.

nonsense

[timmarhy]

we aren't even close to the processing power of the human brain.

i'll make a prediction of my own - this guy is after funding.

Re:nonsense

[bnenning]

we aren't even close to the processing power of the human brain.

We aren't that far off. Estimates for the computational power of the human brain are around 10**16 operations per second. Supercomputers today do roughly 10**14, and Moore's Law increases the exponent by 1 every 5 years. Even if we have to simulate the brain's neurons by brute force and the simulation has 99% overhead, we'll be there in 20 years. (Assuming Moore's Law doesn't hit physical limits).

Re:nonsense

[shura57]

You can similarly compare the temperature of the human brain and then observe that the machines have long bypassed it. Does it make machines smarter? I don't think so.

The brain is so insanely parallel and the neurons are not just digital gates, more like computers in themselves. The machines of today are a far cry from the brain in how they are built. But sure, you can compare them by some meaningless parameter to say that we're close. How about the clock frequency: neurons are 1kHz devices, and modern CPUs are in GHz now...

Re:nonsense

[timmarhy]

it's not just the raw processing power we lack, but the programming language to handle such massively parallel processes.

think of the millions of processes being controlled by the brain at any given moment, on a practical level we lack an OS that could even compare.

each neuron isn't like a transistor, it's more like a cpu on it's own and there are billions of them in a brain. i'm not saying we won't get there at some point, but i do think it's damn site further off then a few decades.

Re:nonsense

[hawkfish]

I'm not even convinced that the universe is deterministic, but current computers most definitely are. The brain may well be taking advantage of physics we don't even know about.

Re:nonsense

[k.ovaska]

I'm not even convinced that the universe is deterministic, but current computers most definitely are. The brain may well be taking advantage of physics we don't even know about.

If universe would not be deterministic, how could computers be? Are they operating in a deterministic subset of a non-deterministic universe?

In fact, computers are not deterministic, if we define determinism as "we are (easily) able to predict computer behaviour". A bit can flip its state erroneously. This happens more easily when CPU temperature is high, for instance. But this is usually seen as a disadvantage.

About brains using unknown physics. The human brain is evolved from more primitive brains. Mouse, for instance, is quite similar to human on genomic level. Did brains of more primitive life forms, which are hardly intelligent, also use yet-unknown physics or did brain cells dramatically change in evolution to take advantage of new physics?

Garbage.

[FooAtWFU]

I have little doubt we already have the components necessary to simulate a human-like brain, one way or another, right now. But I that's not enough. You need to know how to put it together, how to set it up to be educated somewhat-like-a-human, how to get it some semblance of human-like sensory input (at least for vision/hearing centers, if you're interested in either of those things) and then you need to train it for years and years. So, 21-years-off is too optimistic, I think, by at least an order of magnitude, and possibly two or even three.

The "intelligent nanobot" bit is complete and utter garbage, though. Especially by 2029. I don't think you can even begin to fit something "intelligent" in a package about the size of a cell. Even if it's theoretically possible, our technology can currently construct nano-gears one atom at a time.

Don't do it!

[magarity]

(most) People can go out to get more education to advance from a menial job to a more skilled one when taken over by a robot but wtf do we do if the machines are as smart as we are? Who is going to hire any people to do even the most advanced thinking jobs when the machine that works for electricity 24/7 can do it? This kind of thing will bring on the luddite revolution in a hurry.

Re:Don't do it!

[andy314159pi]

wtf do we do if the machines are as smart as we are? Who is going to hire any people to do even the most advanced thinking jobs when the machine that works for electricity 24/7 can do it?

We can raise the price of electricity and lower the cost of candles.

Re:Don't do it!

[lee1026]

alternatively, we can just lazily sit around and get the computers to do all of the work.

Re:Don't do it!

[KKlaus]

Better mail university professors some bombs eh?.

All kidding aside, you've just made exactly the (fairly astute IMHO) point Ted Kaczynski made in his manifesto available here Industrial Society and It's Future.

Pretty good read, even if what he decided to do about it was entirely deplorable.

Re:Don't do it!

[urcreepyneighbor]

My future sexbot won't be putting anyone out of work. ;(

Re:Don't do it!

[noidentity]

Hopefully we can culturally evolve to see the obvious fact that if we have machines doing most of the work, WE DON'T NEED TO WORK.

Re:Don't do it!

[dokebi]

Ah, but you forget the upside. When the machines start thinking for us, we can finally give up this civilization bs and just go back to doing what we've been evolved to do in the first place, mostly make babies and hunt game. Humanity would have finally achieved universal peace and happiness, no matter what some movie might tell you. And those of you who wants to reach for other goals like exploration and knowledge can be free to do so. The most of humanity will just accept the program, and be at peace.

Re:Don't do it!

[freyyr890]

And when the machines break down, or perhaps are destroyed by an enterprising terrorist group who manage to build some sort of EMP device, such as a nuclear weapon in the upper atmosphere?

Suddenly you've got a lot of helpless people on your hands.

Re:Don't do it!

[Anonymous Coward]

With merit being skill and merit being decoupled from wealth in absolute terms. who decides who gets what and why should someone have more then someone else?

Re:Don't do it!

[hawkfish]

With merit being skill and merit being decoupled from wealth in absolute terms. who decides who gets what and why should someone have more then someone else?

Probably the same people that do it now - the owners of the machines.

Re:Don't do it!

[Twinbee]

It's simple, we do what we were meant to do - compose music & create art, socialize more, and generally enjoy life. Eventually pursue exploration of the universe. Sounds good to me.

I bet these hyper-interlligent robots wouldn't have the creative insight into scientific investigation that people would have anyway.

Re:Don't do it!

[lee1026]

when there is no scarcity, who gets what is a purely academic exercise.

Retarded

[mosb1000]

I think these nonsense predications are best described as retarded. You can't predict something that is beyond our current technological capability, since it depends on breakthroughs being made that are impossible to predict. These breakthroughs could come tomorrow, or they could never come at all. I don't know why I'm posting this. Even talking about this fantastic nonsense is a waste of time.

Quite simply BS

[wanax]

Until we figure out how a water buffalo can be an individual at one spatial scale, and part of a herd as a texture at another scale... just in vision... we won't have smart computers.

Fat chance

[popmaker]

For something as inexplicably complex as our brain... which by the way we HAVEN'T understood a fraction of. For something as mysterious as emotions - a problem which lies behind a philosophical problem which we don't KNOW if we ever came closer to solving in the last 2000 years... I say FAT F***ing CHANCE!

Either they are stupid (in which case, I have to admit, solving the problem is a tiny bit easier) or this is publicity

I don't care if there is a while until 2029 hits us, we have nothing real to believe this prediction is useful.

Whatever Could They Mean?

[flyneye]

" Artificial Intelligence will reach the level of humans"
Buddy,I've been around more than four decades.I've yet to see more than a superficial level of intelligence in humans.
Send your coders back to the drawing board with a loftier goal.

Re:Whatever Could They Mean?

[Anonymous Coward]

It's ironic that this is modded 'funny' I think, simply because it's true. Humans have a lot of survival instinct and the like which is makes for a lot of nonsense in modern society.

Artificial intelligence indeed needs to strive for something more appropriate for our needs. Perhaps we should make a super engineer...

Re:Whatever Could They Mean?

[jalet]

Well, if you've seen Idiocracy then they may be right, but it's just the other way around : Human intelligence will reach the level of toasters.

Which human?

[Anonymous Coward]

I'm pretty sure computers are already at the level of intelligence of many prominent humans.

The End of Intelligent Design

[denoir]

It is not too much of an overstatement to say that the field of AI has not significantly progressed since the 1980's. The advancements have been largely superficial with better and more efficient algorithms being created but without any major insights and much less a road map for the future. While methods that originated as AI research are more common in real-world applications, the research and development of new concepts has made a grinding halt - not that it was ever a question of smooth continuous progress.

It might seem like the lack of AI development is a temporary problem and altogether a peripheral issue. It is however neither - it is a fundamental problem and it affects all software development.

Early in the history of computing, software and hardware development progressed at a similar pace. Today there is a giant and growing gap between the rate of hardware improvements and software improvements. As most people involved in the study of the field of software engineering are aware of, software development is in a deep crisis.

The problem can be summarized in one word: complexity. The approach to building software has largely been based on traditional engineering principles and approaches. Traditional engineering projects never reached the level of complexity that software projects have. As it turns out humans are not very good at handling and predicting complex system.

A good example of the problems facing software developers is Microsoft's new operating system Windows Vista. It took half a decade to build and cost nearly 10 billion dollars. At two orders of magnitude higher costs than the previous incarnation it featured relatively minor improvements - almost every single new radical feature (such as a new file system) that was originally planned was abandoned. The reason for this is that the complexity of the code base had become unmanageable. Adequate testing and quality assurance proved to be impossible and the development cycle became painfully slow. Not even Microsoft with its virtually unlimited resources could handle it.

At this point, it is important to note that this remains an unsolved problem. It would have not been solved by a better structured development process or directly by better computer hardware. The number of free variables in such a system are simply too great to be handled manually. A structured process and standardized information transfer protocols won't do much good either. Complexity is not just a quantitative problem but at a certain level you'll get emergent phenomena in the system.

Sadly artificial intelligence research which is supposed to be the vanguard of software development is facing the same problems. Although complexity is not (yet) the primary problem there manual design has proved very inefficient. While there are clever ideas that move the field forward on occasion there is nothing to match the relentless progress of computer hardware. There exists no systematic recipe for progress.

Software engineering is intelligent design and AI is no exception. The fundamental idea persists that it takes a clever mind to produce a good design. The view, that it takes a very intelligent thing to design a less intelligent thing is deeply entrenched on every level. This clearly pre-Darwinian view of design isn't based on some form of dogma, but a pragmatism and common sense that aren't challenged where they should be. While intelligent design was a good approach while software was trivial enough to be manageable, it should have become blindingly obvious that it was an untenable approach in the long run. There are approaches that take the meta level - neural networks, genetic algorithms etc, but it is thoroughly insufficient. All these algorithms are still results of intelligent design.

So what Darwinian lessons should we have learned?

We have learned that a simple, dumb optimization algorithm can produce very clever designs. The important insight is that intelligence can be traded for time. In a short in

Re:The End of Intelligent Design

[scruffy]

The parent is a great post, but I would substitute the 90's for the 80's because that is when machine learning and probabilistic reasoning got themselves mostly straightened out.

But this decade, can anyone provide any significant breakthrough in AI? It seems that any real results are because of faster and more processors and faster and more memory.

Here is my challenge. Can anyone name an important algorithm or representation from this decade?

Re:The End of Intelligent Design

[grcumb]

The problem can be summarized in one word: complexity. The approach to building software has largely been based on traditional engineering principles and approaches. Traditional engineering projects never reached the level of complexity that software projects have. As it turns out humans are not very good at handling and predicting complex system.

I was about to say that I couldn't agree more, but the truth is, I could.

When you say 'complexity', I believe you're speaking about design complexity. There are, however, any number of complex systems at work within the world of technology. You're absolutely right that they are not monolithic, designed systems, though. And your point that AI researchers will continue to bang their heads against this wall is well made.

A good example of the problems facing software developers is Microsoft's new operating system Windows Vista. It took half a decade to build and cost nearly 10 billion dollars. At two orders of magnitude higher costs than the previous incarnation it featured relatively minor improvements - almost every single new radical feature (such as a new file system) that was originally planned was abandoned. The reason for this is that the complexity of the code base had become unmanageable. Adequate testing and quality assurance proved to be impossible and the development cycle became painfully slow. Not even Microsoft with its virtually unlimited resources could handle it.

And yet, there are computer operating environments of equal (and, in some cases, greater) complexity that are thriving and healthy, adapting and even - dare I say it - evolving at a remarkable rate.

How many technical staff does Canonical have, do you think? In spite of being extremely few in numbers, they manage to produce a remarkably complex operating system, upgraded every six months. Of course, we all understand how this is possible. They rely on simple, organic human systems to resolve immensely complex issues. I find it fascinating that it works at all, and would likely argue against the premise if I weren't faced daily with the outputs of these systems.

I think that AI is perfectly achievable, but probably not in ways that most AI researchers might desire. I think it's possible to build extremely complex organic systems with very simple rule sets. I do not, however, think it's reasonable to expect to be able to operate these systems in a jar, as it were. The limitations of classical logic and computer hardware make the emulation of such a system effectively impossible. As others have rightly pointed out earlier in this discussion, that's not going to change.

Nonetheless, AI will become commonplace before too very long. We probably won't recognise it as anything new or unusual, though. It will consist largely of human inputs and interactions, after all.

Re:The End of Intelligent Design

[Jeff DeMaagd]

"And yet, there are computer operating environments of equal (and, in some cases, greater) complexity that are thriving and healthy, adapting and even - dare I say it - evolving at a remarkable rate."

Despite all that, it's not self-replicating. Over a billion transistors in a Core 2 Quad and very little of its circuitry is changeable after it's shipped.

Re:The End of Intelligent Design

[hughk]

First, I believe that you are using a bad example. MS as a company was always run by their marketing needs rather than engineering, hence their problems delivering major systems. The company lacks rigor and discipline in their design process but others do have that and deliver complex software, albeit at perhaps a higher price. The classic method of engineering is to take a problem and reduce it to a series of smaller ones that can be clearly and carefully abstracted. This disciplined approach is how we build bridges and space shuttles and the overall constraint is the design. For evolved systems, the constraint is the environment, anything can be tried with poorer results being culled by being outcompeted or even culled. It is clear that the brain has a structure, but it appears to far from a classic, top down structure.

Re:The End of Intelligent Design

[Animats]

Can anyone name an important algorithm or representation from this decade?

There's been substantial progress in trainable computer vision systems in the last decade. Computer vision is finally starting to work on real-world scenes. SLAM algorithms work now. Texture matchers work. There really has been progress in those areas.

Re:The End of Intelligent Design

[master_p]

The reason Vista cost so much but has nothing special to offer is because Microsoft does not know what 'separation of concerns' means. Windows is a big ball of mud.

Perhaps the problem of AI is approached in the wrong way. Perhaps the only mechanism that is required for AI is an efficient pattern matching algorithm.

I think the brain does not have a complex algorithm. All it does is match the input to responses and recall those responses. All of the brain's power goes into parallel searching, something that computers are not efficient in doing.

Re:The End of Intelligent Design

[k.ovaska]

Over a billion transistors in a Core 2 Quad and very little of its circuitry is changeable after it's shipped.

CPUs are universal Turing machines which means they can be programmed to do anything a Turing machine can do. Changing a CPU wouldn't affect its fundamental expressive power. In this sense, CPUs provide a very flexible platform. Changing a CPU might affect its quality attributes, though, like fix defects and improve performance.

Re:The End of Intelligent Design

[Anonymous Coward]

I detect superhuman level intelligence in this post. HAL, is that you?

intelligence is overrated

[Anonymous Coward]

It's been my observation that what passes for so called intelligence is highly overrated.
Haven't you notice how self proscribed smart people are generally the least functional folks (e.g, lack EI or street smarts or are just really dumb).

I for one don't look forward to computers that act like rainman...

I don't even think we (as a scientific collective) even know what intelligence is....

I swear to $Deity

[Anonymous Coward]

If you guys start the A.I. robot revolution before I lose my virginity, I'm totally haunting your asses when I die and you become slaves to the machines!

Kurzweil has no credibility

[nguy]

The guy talks and sells a lot, but he has contributed almost nothing to Artificial Intelligence. Have a look at his publications:

http://scholar.google.com/scholar?as_sauthors=r-kurzweil&as_subj=eng

Re:Kurzweil has no credibility

[NewbieProgrammerMan]

I have to say I don't know enough about AI or CS to know what to make of that list. I hear this guy frequently billed as some uber-genius AI scientist...is that just a bunch of self-promoting-nonsense? Can anybody that actually works in AI comment on him?

Re:Kurzweil has no credibility

[nguy]

I hear this guy frequently billed as some uber-genius AI scientist...is that just a bunch of self-promoting-nonsense?

Yes.

Kurzweil has founded a bunch of moderately successful companies that have sold software vaguely related to AI.

I have to say I don't know enough about AI or CS to know what to make of that list

You don't really have to know that much; just compare what he has done to some other AI people, by searching on Scholar for Peter Norvig, Patrick Winston, Rodney Brooks, Tom Mitchell, or Terry Winograd.

Sounds Familiar?

[zdude255]

Didn't they predict the same thing 21 years ago?

Don't think so

[DreadPiratePizz]

Predictions like this have been made in past, and not even come close. This one is no different. The bottom line is that humans process some information in a non-representational way, while computers must operate representationally. So even if the computation theory of mind is true, a microchip can't mimick it. Hubert Dreyfus has wrote a great deal on this topic, and provides extremely compelling arguments as to why we'll never have human type AI. Of course, AI can do a lot of "smart" things and be extremely sophisticated, but it will never pass am unrestricted turring test.

Re:Don't think so

[bnenning]

Predictions like this have been made in past, and not even come close. This one is no different.

The difference is that in 20 years we may have sufficiently powerful hardware that the software can be "dumb", that is, just simulating the entire physical brain.

The bottom line is that humans process some information in a non-representational way, while computers must operate representationally.

What prevents a computer from emulating this "non-representational" processing? Or is the human brain not subject to the laws of physics?

Re:Don't think so

[radimvice]

Hubert Dreyfus has wrote a great deal on this topic, and provides extremely compelling arguments as to why we'll never have human type AI.

Compelling arguments, perhaps, but only to your supple, human brain and its weakness for non-representational information. I'm sure the computers would disagree.

Re:Don't think so

[DreadPiratePizz]

What prevents a computer from emulating this "non-representational" processing? Or is the human brain not subject to the laws of physics?

Because computers MUST use representation in order to compute. Those 1s and 0s have to stand for something.

Re:Don't think so

[glwtta]

The difference is that in 20 years we may have sufficiently powerful hardware that the software can be "dumb", that is, just simulating the entire physical brain.

You expect to be able to simulate roughly 1.4 x 10^26 atoms in 20 years? We can't simulate one today!

Don't get me wrong, I love the can-do attitude that producing intelligence is basically a matter of ordering a couple of extra blade racks from Dell, but I can't help feeling that the scope of the problem tends to get underestimated around here.

Re:Don't think so

[Anonymous Coward]

>The difference is that in 20 years we may have sufficiently powerful hardware that the software can be "dumb", that is, just simulating the entire physical brain.

Yes, now all we need is a model of the brain. Err, you're aware that there aren't any accurate models of a cockroach, yet?

Re:Don't think so

[Just Some Guy]

You expect to be able to simulate roughly 1.4 x 10^26 atoms in 20 years?

The thinking is that you'd simulate it's structures, not each individual atom. For example, you can model the behavior of a transistor without simulating all the atoms inside it. In this hypothesis, the real question is what level of detail is required to make it work. Kurzweil et al believe that in 20 years, we'll have the computing power and that knowledge and will be able to put them together.

I bought "The Singularity Is Near" on a lark and admit that it sounded preposterous at first. However, there are about 200 pages of footnotes and references backing up every single claim he makes. After a couple of chapters, my opinion shifted from "ridiculous" to "inevitable".

Re:Don't think so

[Anonymous Coward]

1. An "empty" brain is worthless in and on itself.
2. You still need to write the software to actually simulate the brain, which just may be harder than you think.

Re:Don't think so

[roman_mir]

our hardware is not going in the direction of becoming a brain, it really is still the same old adder, only it adds faster.

Re:Don't think so

[BCSWowbagger]

What prevents a computer from emulating this "non-representational" processing? Or is the human brain not subject to the laws of physics?

This sort of rhetorical question presupposes that "minds are simply what brains do," which remains to this day (outside the /. world and certain parts of the ivory tower) a hotly contested claim with strong evidence on both sides. Indeed, this entire conversation only reinforces the point that all Strong AI research is fundamentally rooted to a very particular set of philisophical claims that are espoused by a very particular set of researchers (Minsky et. al.) who, while perhaps very good neuroscientists and cognitive researchers, are terrible at doing philosophy.

Until the basic premises on which Strong AI is built are rigorously examined from a strictly neutral standpoint, the field is just going to keep spinning its stone wheels.

Re:Don't think so

[Anonymous Coward]

Do you really think that emulated physics can surpass real physics in terms of speed, size, and/or power? Just imagine how much smaller, faster, and more powerful than a human brain such hardware would have to be, if you want to run software on it which simulates the physics of a brain that is better than ours.

So at least it would have to be done completely in hardware to even come close to achieving its goal, but even in that case, there would still be digital electronic circuits, probably in 2D, to model complex physical and chemical processes in 3D. Not likely to be faster either.

And then, our brains are supposedly not that different from those of our animal ancestors, and as far as I know science has not figured out what the little differences happen to be. What makes you think we will get something resembling a human brain when we try to simulate something that might as well be an animal brain?

Predictions are useless in this case

[The One and Only]

It's one thing to predict when a building project will be finished or when we'll reach a certain level of raw processing power because these things proceed by predictable means. But strong AI requires us to make theoretical advances. Theoretical advances don't proceed like a building project--someone has to have a clever idea, fully develop and understand it himself and convince others of it. And it won't occur to someone all at once, so we'll need incremental advances, all of which will happen unpredictably.

Luckily for all of us Kurzweil is Stone Cold Crazy

[liquiddark]

If you read a Kurzweil book, it's as if he understands hope and has no concept of problems. The man is so good at glossing over difficulties he should patent his methods and join the magazine industry.

Where is the proof of possibility

[iamacat]

Human brain is non-deterministic and works very differently from computers. We experience self-awareness with it's numerous sensations which is certainly related to brain's electrical activity but is nevertheless not fundamentally explained by the same. Who is to say that this 2029 computer will actually have a consciousness, or that it will have a consciousness similar to ours. On the mechanical level, will it match human senses of touch and smell or is the "baby" computer supposed to develop by looking at world with a single webcam.

I say it's a baseless claim until these questions are addressed. It may be meaningful to say that a computer of a given era will be able to perform a complex but deterministic task which is currently the domain of humans, such as, say, driving a car.

Re:Where is the proof of possibility

[DigitalisAkujin]

Proof of possibility: you.

Brain works on electricity. We know this. We can of course emulate that behavior.

Re:Where is the proof of possibility

[iamacat]

Yes, and my refrigerator also runs on electricity. I should fill in a social security number application and claim it as a dependent on tax return.

Re:Where is the proof of possibility

[piojo]

The human brain and consciousness are complex. We don't know that they are non-deterministic. And furthermore, even if it's fundamentally random on some level, can't that still be approximated with a random (on some level) algorithm? There may be other arguments as to why the brain can't be modeled ("maybe if the brain were modeled as an algorithm, it would have to be *infinitely long*"), but I don't know many / I'm not sure how I feel about them.

Consciousness is also a strange beast. What is consciousness? Why does consciousness feel continuous, when we know it isn't? (Some people even regain consciousness after they have been pronounced dead.) Why do I still think I am the same being that I was 10 years ago, when my brain was made of completely different cells? Because of the uncertainty of these questions, I think that *what consciousness is* really doesn't matter.

Consciousness may just be part of the noise that results when a thinking being becomes self aware. But no matter what it is, I think it developed as a means to an end, rather than an end in itself. If this is so, when we create computers that can parse written information and communicate effectively, it won't really matter whether they are "conscious", and it won't matter what it would mean for such a machine to be conscious.

Re:Where is the proof of possibility

[iamacat]

If this is so, when we create computers that can parse written information and communicate effectively, it won't really matter whether they are "conscious", and it won't matter what it would mean for such a machine to be conscious. Actually, it's THE most important question about such a computer. If it's not conscious, I should feel free to use this computer as a sex slave, arbitrarily examine or modify it's memory or intentionally expose it to damaging viruses. If it is in fact sentiment, even pulling out the power plug from a computer that doesn't boot but is making some intermittent beeps requires a debate on weather it is CPU-dead or in a minimally conscious state.

I happen to believe that consciousness MANIFESTS itself through quantum events and the brain somehow has unique ability to amplify such effects into macroscopic actions. In this case, simple pseudo-random number generators or even most physical sources of randomness would not suffice. But I wouldn't want to be proven wrong and create a paraplegic being destines to end up in a garbage dump in China within 3 years.

Re:Where is the proof of possibility

[Anonymous Coward]

Why did you write that and why am i replying? I think this reply is the culmination of a life time of fairly random events. To actually choose to reply only took a small subset of events, although having a long term interest in how things work, probably influenced me.

I don't know how a brain works, but I think that every input will have some form of output which in itself is an output and will cascade and may eventually show what we think of as intelligence. I think that perhaps people are working on simulating the higher levels of intelligence without taking into account the basics.

we probably have enough sensors to be able to start to simulate simple brains, maybe even complex brains. Just because some sensors are not available doesn't mean a lack of intelligence blind and deaf people still function as intelligent people so do people with extremely limited motor control.

  Dumb devices such as a roomba can still perform some quite seemingly intelligent actions. I am thinking of how they manage to recharge. The roomba has sensors that tell it how much power it has available and at a particular power level it will begin to seek out its charging base. It isn't that clever but its relatively clever behaviour. If you supplied the charging base with say solar power then perhaps your roomba has some form of life, obviously a very restricted form of life, although being a roomba it will die fairly quickly.

So is a roomba intelligent, probably not enough, It can't learn anything, thou we have some software that learns ocr and speech recognition programs do a good approximation of learning. only the learning is controlled by the code we put into it.

Suppose we mounted a laptop on a roomba and gave it a speech program. Perhaps we could instruct it how to do some new actions, even better would be if we let it communicate with other roomba-laptops

initially i was thinking use wifi to let them communicate what they have learnt but i think speech would be better as it gives a chance of imperfect communication Theres a possibility it would learn a unique action

The more roombas you add to this community the more likely it is that individual roombas would have different skills...

just a thought

But do we really need it?

[3seas]

As someone once wrote "Artificial Intelligence - nothing is naturally that stupid."

But on another note, what is Artificial Intelligence anyway, but the by-product illusion of Automating Information (static, active and dynamic) enough to create the illusion?
What all is involved in Automating Information but simply applying what we already do in creating and dealing with abstractions, but through a hard mineral based computer instead of living biological tissue known as the brain.

Then there is another perspective, with the amount of artificially intelligent people we have running around, do we really need or want machines to emulate them?

Re:But do we really need it?

[aim2future]

Then there is another perspective, with the amount of artificially intelligent people we have running around, do we really need or want machines to emulate them?

It depends on the type of society you want to live in. Any type of utopian society needs AI for moderation and assistance.

We as humans are too unreliable to be trusted with lots of power and we are unable to keep any type of utopian experiment running for long.

The challenge I see is not so much about creating the intelligence, but creating reliable, ethical trusty intelligence with love. We probably don't want to create super intelligent liers who are doing stuff in their self interest.

Re:But do we really need it?

[3seas]

The idea of a utopian society is not a realistic idea, as we all seem to have different ideas as to what that would be, meaning someone will be unhappy and consider resistance to the utopia and as such voiding it.

The idea that we can create something better than us is an illusion, though we may be able to take the best of us and incorporate it into a machine intelligence there is no proof that summing what is known to be imperfect will result in the perfect. And this is the problem. Though we may be able to create something that seems perfect this way, the reality is that its flaw is likely of such a high level that we won't see it until it to late.

There is also a matter of war mentality where the proof is the insulting amount of money budgeted for military defense worldwide in speculation of needing protection vs. the amount of money needed to genuinely address real world issues which removes war and terrorism motives. See: What the World Wants noting that of the 6 plus billion people on this planet its a fraction of 1% that cause such "maybe needed" spending while ignoring the more certain direction of removing motive of warfare.

Waring mindsets get first crack at any advances in technology to see how it can be used for war and defense against.

To know the motives of 9/11 read Trillion dollar bet and understand the probability of those who will be inconsiderate of the needs of others in their goal of being better than others, financially.

We May see skynet or wopr at that time

[Joe The Dragon]

and it will start a Global Thermonuclear War

Mod parent UP!!!

[mcrbids]

Man, oh man, I wish I had some mod points right about now!

This has to be one of the most insightful posts I've read in a long time - this subject could easily be expanded into a book and I'd buy the book!

Re:Mod parent UP!!!

[Jugalator]

Yeah, I hadn't even really thought about this that way, and it seems quite right to me as a developer too. We no longer have the revolutions in software we had early on. Windows Vista is scarily similar to Windows 95. But Windows 3.1 wasn't to MS-DOS. So it's not just a problem of looking at the wrong timeframes. The timeframes are 11 years in both cases.

Re:Mod parent UP!!!

[Lord Ender]

Hm? This is your standard sales pitch for genetic algorithms... ended with some vague call for "recursive" genetic algorithms... It's nicely worded, but there is no new value in it unless he explains what a recursive genetic algorithm is in some meaningful way.

You've underestimate Apple

[Anonymous Coward]

In 25 years, the iAI that autonomously walks (with a strut) will be introduced by Apple. It will not play chess, however it will play checkers, and the board will be setup perfectly.

Re:You've underestimate Apple

[HeroreV]

If Apple went anywhere near the subject of androids, they'd make something like iBimbo. It'd be beautiful, and maybe even have a wonderful voice, but it would have extremely low intelligence.

Re:You've underestimate Apple

[corbettw]

If Apple went anywhere near the subject of androids, they'd make something like iBimbo. It'd be beautiful, and maybe even have a wonderful voice, but it would have extremely low intelligence. How do I get on the announcement list for this?

Re:You've underestimate Apple

[HeroreV]

Just visit digg.com a lot. You'll see at least 80 submissions a day on anything Apple related.

Too bad he likes to screw the taxpayer.

[Anonymous Coward]

Someone who charges this much for educational software is more concerned with helping his wallet then helping students. Kurzweil 3000 for Windows Professional Color Windows-based reading, writing and learning software for struggling students. 4 Details $1,495.00

How pathetic

[JulianConrad]

It's pretty clear now that the rate of progress has leveled off and fallen well short of the flying cars, space colonization, nanotech assemblers and friendly AI fantasy-future. Physicist Jonathan Huebner has gathered empirical evidence (PDF) showing that we're pretty much fucked for new, practical technological ideas already, and that includes AI. I'd respect Kurzweil more if he'd stop making an ass of himself with his sci-fi stuff, go back to his lab and work on something useful.

Re:How pathetic

[LaskoVortex]

I'm not going to certify myself as a lunatic by arguing that we will get the "AI fantasy-future" and the other things you mentioned any time before the sun burns out, but I just read the Heubner paper and his metric for innovation is patently weak. First, he is using a metric based on culling "technology events" from a non-peer reviewed book of written by a third party (Hellemans and Bunch). Second, he is using only one metric instead of several independent metrics that one could compare. Stronger would be to assemble measurements from independent peer reviewed publications, or to collect measurements himself. He would also need to develop a more compelling definition of the term "technological event" (more compelling, that is, than "Hellemans and Bunch said so"), and to compare these now well-defined "technological events" across fields (e.g. manufacturing, engineering, physical sciences, life sciences). We could then assess the validity of this definition by comparing numbers to our expectations. For example, we expect a recent spike in technological events in the biological sciences. Were I a reviewer for this paper, it would have a tough time making it into any journal in its present form. Notice also that Heubner doesn't even have an affiliation with any academic institution.

Re:Luckily for all of us Kurzweil is Stone Cold Cr

[Anonymous Coward]

If you read a Kurzweil book, it's as if he understands hope and has no concept of problems. The man is so good at glossing over difficulties he should patent his methods and join the magazine industry.

Indeed, I wonder why he doesn't have a job at SCO...

I agree...

[Jane Q. Public]

As an party "outside" the field but interested, I agree with all of you here so far, except that of course you disagree on timelines. :o)

"Artificial Intelligence" in the last few decades has been a model of failure. The greatest hope during that time, neural nets, have gone virtually nowhere. Yes, they are good at learning, but they have only been good at learning exactly what they are taught, and not at all at putting it all together. Until something like that can be achieved (a "meta-awareness" of the data), they will remain little more than automated libraries. And of course at this time we have no idea how to achieve that.

"Genetic algorithms" have enormous potential for solving problems. Just for example, recently a genetic algorithm improved on something that humans had not improved in over 40 years... the Quicksort algorithm. We now have an improved Quicksort that is only marginally larger in code size, but runs consistently faster on datasets that are appropriate for Quicksort in the first place.

But genetic algorithms are not intelligent, either. In fact, they are something of the opposite: they must be carefully designed for very specific purposes, require constant supervision, and achieve their results through the application of "brute force" (i.e., pure trial and error).

I will start believing that something like this will happen in the near future, only when I see something that actually impresses me in terms of some kind of autonomous intelligence... even a little bit. So far, no go. Even those devices that were touted as being "as intelligent as a cockroach" are not. If one actually were, I might be marginally impressed.

Re:I agree...

[TempeTerra]

We now have an improved Quicksort that is only marginally larger in code size, but runs consistently faster on datasets that are appropriate for Quicksort in the first place.

Can you give me a link to that? I don't keep up with the literature but I'd love to read more, and Google isn't being helpful.

Re:I agree...

[CryoPenguin]

Probably not what the OP was thinking of, but I did find some related literature. 2nd hit on google, actually:
Optimizing Sorting with Genetic Algorithms, by X. Li, M.J. Garzaran, and D. Padua, in Proc. of the International Symposium on Code Generation and Optimization, pages 99-110, March 2005

Re:I agree...

[greyhueofdoubt]

You brought up an interesting phrase there- 'trial and error'. I think it's interesting that you bring that up, because that is precisely how babies/children DO learn before language and cognitive development allow them to absorb spoken instructions; even then, trial and error make up a huge portion of an infant's learning process.

A 'learning' AI would still use brute force/trial and error to achieve tasks early on, but would retain those lessons as experience increased. Much like people and other animals, as the system aged, it would become more skilled. Eventually, I would expect it to become almost dogmatic in its actions, much like humans, and for the same reasons- the 'ruts' formed by constantly reinforced data would drown out minor inconsistencies or new trends to the level of statistical noise.

The silver lining of this is that 1- the system would be machine-readable and writable, making duplication after the initial development trivial; and 2- 'learning' could be halted at a certain point, maybe at a point where the machine reaches its optimal reliability/efficiency, and then duplicated as a well-understood 'production' machine.

Work needs to be done on neural-net-type processing and very fast, very abundant memory in order to make this work, but everyone knows that... I'm just saying.

-b

Re:I agree...

[32771]

I'm not sure of your metric for sorting algorithms, but the STL uses some modified quicksort algorithm.

http://en.wikipedia.org/wiki/Introsort

Wiki seems to have covered sorting algorithms quite well, this is definitely some good starting point.

The development of the intro sort algorithm seems to be human driven. This questions the statement though
that there hasn't been any progress on quicksort in 40 years. The greatest speed this particular algorithm
can be executed at seems to be bounded at O(nlog(n)) though.

The Wiki entry about general sorting algorithms states:

"The following table describes sorting algorithms that are not comparison sorts. As such, they are not
limited by a O(n log n) lower bound."

No Quicksort like algorithm should be able to exceed that bound. Somehow I would like to see a
proof of that now.

Re:I agree...

[32771]

This is very funny I found it at digg! http://www.digg.com/programming/Natural_Selection_Yields_a_Sorting_Algorithm_Better_than_Quicksort Can you click on the link as a self respecting Slashdotter ;?

Not a chance

[Progman3K]

We'll certainly have machines that will appear to think and act human, but self-aware?
Nope, I don't buy it.

It's like building a tower to reach the moon: You are able to double the height of the tower every year for the first n years, so based on the rate of growth, you could calculate that we would be on the moon soon, the only problem is that there are implicit limits to how high the tower can become until it is too heavy or too unstable to continue standing.

Everyone claiming we'll have true AI by 2029 is making the same types of mistaken assumptions.

I think human level AI is possible...

[Anonymous Coward]

I think human level AI is possible, but not in twenty years, and *not* using binary logic. Binary logic is indeed a part of human intelligence, but only a small part. In contrast, binary logic is the *only* type of logic computers are capable of doing. Yes, they are very good at it, which allows programmers make computers do many things that seem beyond the scope of basic binary logic, but that's all they can do.

When real AI is acheived, it won't be a binary computer, it'll be something that hasn't been invented yet.

Re:I think human level AI is possible...

[jnana]

IANAAICS... But the fact that at the hardware level the computer is binary does not matter.

On top of that hardware layer, you can implement, for example, reasoning based on non-monotonic logic, which allows for the very human phenomenon of coming to a tentative conclusion based on incomplete but best available evidence when necessary, and revising those conclusions when new information is available. The fact that it is implemented in something that compiles down to ones and zeros is irrelevant.

Anything that can be represented can be represented with ones and zeros.

Predicted by AI

[halcyon1234]

The grain of salt here is that the date was predicted by a 2008 AI which, as we all know, are not anywhere near as smart as a 2009 AI.

The sacred brain and other myths

[denoir]

This is a sort of continuation of the parent post.

The comedian Emo Philips once remarked that "I used to think my brain was the most important organ in my body until I realized what was telling me this."

We have tendency to use human intelligence as a benchmark and as the ultimate example of intelligence. There is a mystery surrounding consciousness and many people, including prominent philosophers such as Roger Penrose, ardently try to keep it that way.

Given however what we through biological research actually know about the brain and the evolution of it there is essentially no justification for attributing mystical properties to our data processing wetware. Steadily with increased capabilities of brain scanning we have been developing functional models for describing many parts of the brain. For other parts that need still more investigation we do have a picture, even if rough.

The sacred consciousness has not been untouched by this research. Although far from a final understanding we have a fairly good idea, backed by solid empirical evidence that consciousness is a post-processing effect rather than being the first cause of decision. The quantity of desperation can be seen in attempts to explain away the delay between conscious response and the activations of other parts of the brain. Penrose for instance suggests that yes, there is an average 500 ms delay, but that is compensated by quantum effects that are time symmetric - that the brain actually sees into the future, which then is delayed to create a real-time decision process. While this is rejected as absurd by a majority of neuroscientists and physicists, it is a good example of how passionately some people feel about the role of the brain. It is however painstakingly clear that just like we were forced to abandon an Earth-centered universe we do need to abandon the myth of the special place of human consciousness. The important point here is that once we rid ourselves of the self-imposed veil of mystery of human intelligence we can have a sober view on what artificial intelligence could be. The brain has developed through an evolutionary optimization process and while getting a lot of benefits it has taken the full blow of the limitations and problems with this process and also its context.

Evolution through natural selection is far from the best optimizing method imaginable. One major problem with it is that it is a so called "greedy" algorithm - it does not have any look ahead or planning capabilities. Every improvement, every payoff needs to be immediate. This creates systems that carry a lot of historical baggage - an improvement isn't made as a stand-alone feature but as a continuation of the previous state. It is not a coincidence that a brain cell is a cell like any other - nucleus and all. Nor is it a cell because it is the optimal structure for information processing. It was what could be done by modifying the existing wetware. It is not hard to imagine how that structure could be improved upon if not limited by the biological building blocks that were available to the genetic machinery.

Another point worth making is that our brains are optimized not for the modern type of information processing that humans engage in - such as writing software for instance. Humans have changed little in the last 50,000 years in terms of intellectual capacity but our societies have changed greatly. Our technological progress is a side effect of the capabilities we evolved that increased survivability when we roamed the plains of Africa in small family hunter-gatherer groups. To assume the resulting information processing system (the brain) would the ultimately optimal solution for anything else is not justifiable.

There has been since the 1950's ongoing research to create biologically inspired computer algorithms and methods. Some of the research has been very successful with simplified models that actually did do something useful (artificial neural networks for instance). Progress has however been agonizi

Re:The sacred brain and other myths

[Anonymous Coward]

Not sure if intelligent design in software has run its course, yet. But as you pointed out, consider the source of that argument. I myself have a vested interest in keeping that method alive. Regardless of that, I have previously done a little thinking about non-intelligent design as an intellectual diversion for myself. I've tried to consider computational equivalents to the core elements of biological evolution and ran into a roadblock.

Warning: novice of cellular biology talking here.

1. The protein seems to be the main agent of molecular activity in the cell. Random errors in DNA cause alterations in constructed proteins. Proteins are pretty remarkable in that they have elements in their chains that are hydrophobic or hydrophylic (or neither) and that causes sections to knot together or expand apart when immersed in water. This gives them a three dimensional shape, which when combined with molecular vibration allows them to act as 3-dimensional machines that act upon other molecules (other proteins or non-proteins). And I guess it obvious that these interactions are chemical and subject to the physical rules of chemistry. So what you need for a computation equivalent would be something like software components which interacted in a codified way like with the rules of chemistry, plus something like the protein which could yield new (non-preprogrammed) operations upon the existing components of the system when subjected to random rearrangement.

2. There is also the problem of the program interacting with its environment. Early biological lifeforms were embedded in a medium (the ocean) that brought them in contact with other lifeforms and other environmental elements until they developed senses and motive means to take charge for themselves. And once again, it is the physical rules of chemistry that control the interaction. If you wanted to evolve an operating system what is the medium that brings the lifeform (the program) in contact with the environment (computer storage, i/o, etc.). Also, once again it seems there is a need to have the computer environment be able to interact in a way analogous to physical chemistry.

Just want to share those thoughts in case they might be useful in your pursuit. Good luck.

Re:The sacred brain and other myths

[Anonymous Coward]

>> As conventional software development through intelligent design is already now unsustainable, the responsibility falls on the AI research. And that field can only be sustainably successful when the mainstream of it understands that it is ultimately software that should recursively build increasingly better software.

It appears like you are blaming software for the human race's short comings, and are also looking at it for a solution.

Perhaps smarter software isn't the answer after all- why are we looking to recreate a resource we theoretically already have? However, how does one go about making smarter people? Breeding programs, novel education procedures and research, greater rewards for technological or mathematical pursuits?

Re:The sacred brain and other myths

[Msdose]

The study of the brain can and will certainly yield additional insights into how to build new algorithms, we shouldn't expect any miracles and we should certainly not have human intelligence as a goal. There is absolutely no reason to end our ambitions there or to even see biological intelligence as a role model for the artificial kind.

I would think we should exhaust the possibilities of the brain's processing program to achieve A.I. Whatever the shortcomings of it's physical substrate and structure, it obviously is capable of doing the job. My guess is that it is simply processing language (words) in a hierarchical manner of some sort (no pun intended). I am astonished that it hasn't been done yet and can only blame the west's slide into communism. No doubt the Chinese will be selling it to us any day now.

Re:The sacred brain and other myths

[hawkfish]

Penrose for instance suggests that yes, there is an average 500 ms delay, but that is compensated by quantum effects that are time symmetric - that the brain actually sees into the future, which then is delayed to create a real-time decision process. While this is rejected as absurd by a majority of neuroscientists and physicists, it is a good example of how passionately some people feel about the role of the brain.

On the other hand, Dean Radin (while barking mad in some ways) has done an experiment that suggests that this is actually correct. His results have been reproduced by other groups using different equipment but the same basic idea: You do a series of tests showing people a mixture of disturbing and relaxing images under computer control and record their responses. Radin use galvanic skin responses, and a European team used live brain MRI (IIRC) but both got the same result - there is a statistically significant incidence of response before the stimulus. I don't know what this says about consciousness, but it seems reasonable to investigate atemporal phenomena in the brain more carefully. John Cramer's Transactional Interpretation of Quantum Mechanics seems relevant here as a physical model for how this might happen. And if Cramer is right, then it seems reasonable that evolution would be able to harness such physical laws even if we don't know about them.

And in the case of anything that challenges deterministic orthodoxy, we may be talking about laws that a large number of working scientists refuse to even consider - even though there is no reason a priori why they should not. The irony here is that determinism grew out of theological notions of Natural Law but is now so much a part of the scientific culture that anything else is treated as heresy!

Re:The sacred brain and other myths

[roman_mir]

We are not going to get anywhere with evolutionary software because the hardware that is running it is still designed and limited to the same old problem of adding two numbers together. Before software can really evolve, the hardware has to evolve and since our hardware is digital it cannot evolve. Digital hardware relies on everything being exact. The timing has to be perfect, the paths of work have to be synchronized.

Our hardware model is useless for evolution. For evolution hardware has to be flexible, exact timing shouldn't be required, instead messaging systems have to be used and signaling should be analog with ranges of options rather than with only two discrete states.

Also our hardware and software together make a total system, hardware is useless without software and our model does not allow the software to modify the hardware. To allow for evolving system the hardware and software must allow for feedback from and to each other.

An evolving computer is analog, that does not depend on a limited number of discrete states and uses messaging rather than perfect timing. It must be self-modifiable on both hardware and software level and hardware must have feedback to and from software and vice versa.

Until we develop a hardware model that satisfies these conditions we will have no useful evolving software systems.

But it does not mean we cannot have computer systems that help us to develop more complex systems, we have such things today. These are specific tools like cad tools for example. With our current hardware model what we can do is develop a multitude of tools, object recognition, speech recognition, image recognition and so on. These tools can be used to define the business problems that we are trying to solve in higher level terms. This is where we can use our current hardware/software models to help us with software complexity.

Re:The sacred brain and other myths

[Fatalis]

The irony here is that determinism grew out of theological notions of Natural Law but is now so much a part of the scientific culture that anything else is treated as heresy!

that's weird, isn't the Copenhagen interpretation almost universally accepted? you know, the one that did away with determinism and showed that the universe is fundamentally stochastic

Re:The sacred brain and other myths

[Raindance]

Certainly the best Slashdot comment I've seen so far in 2008. (The pair of yours, taken together.)

I do have one nitpick. You say,

Another point worth making is that our brains are optimized not for the modern type of information processing that humans engage in - such as writing software for instance. Humans have changed little in the last 50,000 years in terms of intellectual capacity but our societies have changed greatly.

Hawks, Cochran, and a few others have come out with a paper that strongly disagrees with this statement (PNAS, I can dredge it up if you'd like). Their intentionally conservative estimates indicate that selection has been sped up roughly two orders of magnitude in the timeframe you give-- essentially, that culture has provably amplified selection considerably (on some 7% of the genome- with a disproportionate skew towards brain and immune system genes).

I realize, in the grand scheme of things, the amount of cognitive evolution in the past 50,000 years is dwarfed by the amount that has happened previously, so your statement can be taken to be quite correct regardless. But I feel it's important to note that evidence is mounting that we are very cognitively different from even anatomically modern humans of 50,000 years ago, and that even significant cognitive change can probably happen quickly, especially with all the raw material evolution has provided us.

Re:The sacred brain and other myths

[ShooterNeo]

I AGREE with you on the software design part. Our entire model has a MASSIVE, single flaw that I perceive (not quite sure what the 'intelligent design' thing you are talking about is a problem). Simply put, computer code is linear : it is a series of exactly described mathematical operations, where an off by one error ANYWHERE can bring everything else crashing down.

For instance, in windows, one function can't accept an "almost correct" binary answer from a previous function because binary math tends to produce gibberish answers when things go wrong. A single line of code carelessly used can break thousands, even millions of lines of other code that is working correctly.

Neurons are not optimal, but a single misfiring neuron has little or no effect. (in fact, the synapses attached to that neuron would automatically become weaker, limiting the effects of that neuron). "Answers" to a problem consist of a firing pattern coming from an axon trunk, and if that firing pattern is a little bit off from what the next stage of neural processing expects, it still works.

If we could create software that mimicked this, FAR more complex projects would be practical. Instead of subroutines, there would be "modules" of virtual neurons that could be linked together in various ways. These modules would have been "pre-trained", and would have their learning function disabled, so they would act in predictable ways.

So, for instance, you could start with the output from a video camera. Add in a module that converts the camera data into line spots, then a subsequent modules that converts the data into vectors. Link that output to the next stage, and so forth.
(the reason I give this specific example is that we know WHAT the lower stages of visual processing due in mammals)

How to implement computer software this way is unknown. I just know that if neural net "modules" could work, we could make software MUCH more reliable and complex at the same time. It would even be spookily capable of recognizing input it shouldn't. Mistype a command into an operating system, and the text parsing module would probably still know what you meant, because enough virtual neurons fire to 'recognize' the garbled command. And so the OS would still do what you wanted.

From this basis it is not hard to imagine eventually creating an intelligence that pushes things over the cliff. For the 2029 prediction to be correct, we don't actually have to design a human level AI - JUST an AI good enough to improve itself recursively and exponentially. (such that it might start at below "human" intelligence in some respects, but due to enormous processing speed, iteratively improves itself explosively rapidly until it ends up far beyond the human level)

But will it

[EEPROMS]

bother talking to us. An artificial inteligence system could do trillions of calculations by the time a human asks "are you happy ?"

Only 18 Original Thinkers?

[Anonymous Coward]

According to Mr. Boyers, there are only 18 influential thinkers, and Mr. Kurzweil is one of them. One should not simply state as fact such extreme and controversial claims as those, without offering either a reference or an argument in support of them. Please make your case, Mr. Boyers.

That is totally going to

[Lewrker]

spoil 2029 - the year of the linux for the desktop.

Hello? FDA anyone?

[Orleron]

I'm no expert on AI, so for all I know, the technology could reach human intelligence by 2029. But nanobots that crawl through your brain? That I can comment on. Bone Morphogenic Protein (BMP) was discovered by Urist and Reddi in the 1970's, and it took 30 years just to make that product, a simple growth factor, go from bench top to human clinical product. You're telling me that nanobots, a medical device never before seen by the FDA so far, can be approved and ready and in use in humans by then? Let me set the record straight. Even if artificial intelligence reached human level TODAY, there would be no nanobots crawling through our brains by 2029... maybe by 2039 or 2049. Possibly. So whatever year AI reaches human intelligence level, add 30 to 40 years onto that and you'll have your year for a medical product of that magnitude. Remember, the FDA does not care what science and engineering can do, only that they can do it safely and effectively, which is a lot more difficult to show than a simple experiment proving a concept.

AI Insanity

[Anonymous Coward]

Before hooking up an AI at human level or beyond, I really hope they make sure it has human level senses. An AI with no body sensation (sense of self), and none of the five major senses, and no way to learn of its environment, is going to have an extremely unhappy time of it. It would likely be even worse if it's been preloaded with information enough to reason in a human-ish way.

Blue Brain Project

[vikstar]

The blue brain project is already simulating a cluster of 10,000 neurons known as a neucortical column. Althought quite good already (in terms of biological realism), their simulation model is still incomplete with a few more years work to get the neurons working like in real life. With more computational power to increase neuron count and better models they will be able to one day simulate an entire mammalian brain.

Re:Blue Brain Project

[BWJones]

Althought quite good already (in terms of biological realism),

While this project is verrry cool, they are not even remotely close to biological realism. Sorry...

their simulation model is still incomplete with a few more years work to get the neurons working like in real life.

That is just it. We are finding that real biological systems from complete neural reconstructions are far more complex with many more participating "classes" of neurons with much more in the way of nested and recurrent collateral connectivity than is predicted by any existing model of neural connectivity.

Re:Blue Brain Project

[smallfries]

From the reports that they've published I thought their simulation was quite realistic - but I'm a computer scientist rather than a neurobiologist. What are the major differences between their simulation and the real thing?

Re:Blue Brain Project

[BWJones]

First off, they are likely dramatically under representing synaptic connectivity through both conventional synapses and gap junctions. On top of that they are essentially modeling currently understood pathways and are under-predicting the recurrent feedback loops that are apparently the most common type of synapse in neural systems. Finally, definitions of neuronal "classes" are apparently underpredicted as neurons are more than simply connectivity circuits. There is pharmacology and other very subtle physiology at play as well.

What about superhuman hybrid A.I.?

[robotsrule]

Whenever I see stories like this and the usual negative rebuttals that follow, I wonder if I am the only person who read Asimov, Clarke, Crichton, Roddenberry, Heinlein and many others. I am starting to believe that it is because we feel we have "dealt" with the bogeyman of "truly aware" A.I., now that it has been confronted handily by Hollywood via The Terminator and its ilk. In the same way that it was almost comforting to embrace the dark specter of biological terrorism as a pleasant relief from the more real and closer danger of nuclear destruction; focusing on the dawn of A.I. is a relief from the true technological tsunami heading our way.

In the midst of all this talk of pure A.I. is the real steady progress being made in hooking mammalian brains to computers. So far it is in the safe yet icky domain of direct control over robots and other advanced technical based prosthetics, but it is the door to the bigger more powerful scenario that may await us compared to the "birth of A.I." to reference The Matrix. What people fail to understand is that we will make huge progress in this area, much faster than in solely silicon A.I. Why? Because we don't have to understand how the mind works to reap powerful benefits from hybrid A.I. like we do with pure A.I. Neurons by their very nature analyze and adapt to patterns and signals, they just need to be connected and protected.

The most disruptive mind-numbing change heading our way is when human brains can connect with each other over a digital conduit like the Internet. What happens when I can expand my consciousness to be able to maintain far more than the average capacity of 4 to 7 active symbols in my mind, by harnessing the brain capacity of others on a shared peer to peer neuronal network? What powerful meta-consciousness will form when your mind can directly alter a visualization held in real time by another, group dreaming as it were? Or perhaps 10 minds, or a thousand? When we unplug, if we ever do, will we feel as if we woke up from a greater more powerful and majestic dream that evaporates as soon as we disconnect because our minds, by themselves and in comparison, are too tiny to hold the more complex patterns a mind cloud can handle? Perhaps feeling like a butterfly who was dreaming that he was a man, now awake and relegated back to simple thoughts of procreation and feeding, to paraphrase Zen?

In closing, what problems which are now intractable to any single human due to their complexity and scope will fall astonishingly quickly to the power of a million minds focused like a laser on their solution? Please don't take the laser analogy lightly. Right now all of us, and any computer programmer knows this all too well, are recomputing and resolving billions of thought problems which are complete duplicates of each other. What happens when all that duplication is virtually eliminated and our minds in unison all take one small slice of a much larger problem and tear it to pieces? Heaven or hell, you decide, but coming a lot sooner than any of us think.

I'm just some guy

[Anonymous Coward]

But it seems like what you said was entirely obvious, but still a complete guess.
That is, there is NO WAY we'll have computers than can think as humans.

What we may have is the theoretical output capacity of the human brain in CPU power, but that doesn't mean it's artificial intelligence.

In fact, I more than doubt, but laugh at the idea we'll have anything like artificial intelligence in 20 years.

If that were true then basically the world as you know it is about to end and be reborn in about 20 years when machines are capable of human like thought and adaptive problem solving.

The fact is, the most life like AI simulations are pathetic. We may have the terraflops of what we imagine the human mind to output, but as you stated our actual knowledge of how the mind works, no less translating that to hardware/software is far behind.

I think we'll have what we today believe to be the human minds teraflop potential in hardware in 20 years, but it'll be many decades before anything like AI really happens.

How can this guy be so brilliant and think that AI would be here so fast. If todays software is an example of our ability to reach for AI, I think we have more like 200 years before you have AI devices that truly think and solve problems beyond trial and error.

Re:I'm just some guy

[HAWAT.THUFIR]

The first AI program will a virus -- skynet ;) -Thufir

Technological Singularity

[icedcool]

According to Vinge it'll happen before 2030.

We'll see.

Tag: Bullshit

[TheMiddleRoad]

I wish I could add the tag, "Bullshit!" Call me a disciple of Searles, but brains make minds and anything else a wild guess. We don't have the algorithms. I doubt we will for a long, long, long time. We don't have the hardware. We won't have the hardware. We have the barest understanding of how the brain works.

Speaking of dumb predictions

[JulianConrad]

Back in 1978 Robert Anton Wilson predicted that we'd be "immortal" by now. Boy, did he get a serious reality-fucking about a year ago.

Kurzweil's rebuttal from his book...

[doug141]

The Singularity is Near has a rebuttal of your first paragraph. Any sucessful part of AI research spins off into its own well-functioning discipline... optical character recognition, dictation software, text-to-speech, etc... they were sci-fi "AI" in 1980 and now they are working technologies. AI research is the umbrella under which only the unsolved problems still lie, and thus is always undone.

Re:Kurzweil's rebuttal from his book...

[Anonymous Coward]

La la la, I'm Ray Kurzweil, I can't hear you! The singularity will come!! Everything will be taken care of for us!!! I'm gonna live to be 500!!!! La la la!!!!!

Re:Kurzweil's rebuttal from his book...

[Anonymous Coward]

Any sucessful part of AI research spins off into its own well-functioning discipline... optical character recognition, dictation software, text-to-speech, etc... they were sci-fi "AI" in 1980 and now they are working technologies.

No actually that's roughly speaking 1970's technology. Speech recognition 1977 and 2007 is no different when it comes to software. It's all just Hidden Markov Models which were first created in the 1960's. The 1980's brought neural nets, the 1990's some more statistical algorithms such as ICA (independent component analysis) and a slight generalization of neural nets, the support vector machines. This decade the only relevant new AI technology are LSTMs - long short-term memories a minor addition to neural nets that is better at dealing with time series. And what is the cutting edge today? Bayesian belief network models. Those were invented in 1913 with the theoretical framework laid out in the 1770's.

The point is that no major theoretical discovery has been made since modern artificial neural networks were invented in the 1980's (although the basic concepts were developed in the 1950's).

Re:Kurzweil's rebuttal from his book...

[TrekkieGod]

Any sucessful part of AI research spins off into its own well-functioning discipline... optical character recognition, dictation software, text-to-speech, etc...AI research is the umbrella under which only the unsolved problems still lie

The technologies you mentioned are hardly "solved problems." I even hesitate to call them "well-functioning." They sort of work, but there's a lot of room for improvement. OCR is practically worthless for anything that's not a computer font (yeah, new methods for reading captchas appear all the time, but new types of captchas get deployed which need a better OCR to read them in response...all the while, the letters can still be recognized by humans). Dictation software still runs into "Dear Aunt let's so double the killer delete select all" issues and text-to-speech still fails at reading paragraphs with proper inflections so that the text sounds natural.

I'd be ecstatic if just those technologies you mentioned reached human-level status in 20 years. Having a fully functional AI is out of the question. Computers can do many, many things better than humans, but humans are still light years ahead of the most powerful computers in other things. Can you listen to a song and isolate the sounds from a particular instrument? The human voices? Special effects in a movie as different from the music? Can you isolate what one person is saying in the middle of a crowded room by just honing in on their voice and filtering out the others? Can you switch to try to listen to somebody else by picking up their voice?

The human mind is an INCREDIBLE filter and processing computer for all the inputs we get. We're simply not anywhere close to duplicating that functionality.

what will you feed it?

[BorgCopyeditor]

With more computational power to increase neuron count and better models they will be able to one day simulate an entire mammalian brain.

Great. How long will it take to produce good models of sensory input, feedback loops concerning regulation of autonomic systems, and all the rest? Or has no one even begun seriously to work on such things? In any case, add the years necessary to sort that out to your confident predictions.

Re:what will you feed it?

[vikstar]

Great. How long will it take to produce good models of sensory input, feedback loops concerning regulation of autonomic systems, and all the rest? Or has no one even begun seriously to work on such things? In any case, add the years necessary to sort that out to your confident predictions. The idea is not to simulate a living organism, or even an entire nervous system, but to learn more about how the brain works. Download the papers and have a look before dismissing it with the wave of your hand.

Re:what will you feed it?

[BorgCopyeditor]

If anyone is engaged in hand-waving here, it's those who ignore the hard distinction between "simulating a mammalian brain" and "simulating a lump of neurons connected to each other in the ways such neurons might initially be connected in brains." The former requires (at a minimum) the things I mention. The latter may be possible (and may even be useful), but it is by no means simulating a brain.

Re:what will you feed it?

[vikstar]

If anyone is engaged in hand-waving here, it's those who ignore the hard distinction between "simulating a mammalian brain" and "simulating a lump of neurons connected to each other in the ways such neurons might initially be connected in brains." The former requires (at a minimum) the things I mention. The latter may be possible (and may even be useful), but it is by no means simulating a brain. Now you're getting into semantics. If everything you mentioned previously is simulated, then a chemist can come and say "where are the individual molecular interactions?", and if that is simulated one day then a physicist could say "where is the tunneling?" etc.

The problem with research scientists is that many of them believe their level of abstraction is the most important. The point of a simulation is to only simulate those parts that you need for the project. For example, reinforcement learning in the field of artificial intelligence simulates the way a human brain works, in its own level of abstraction, without any ion channels or neurons.

Stating broad issues that need to be addressed for a simulation to fit into your own level of abstraction, if it differs from the project, isn't an insight that is productive to the current research at hand.

Re:what will you feed it?

[BorgCopyeditor]

I'm getting into semantics? I'm not the one claiming to be able to "simulate a mammalian brain." Someone who knows how brains work (and how animals work) would not claim to do such a thing if he could not provide the proper input for this "device." The vastly complicated input necessary for real learning (i.e., sensori-motor feedback loops) simply will not work themselves out, no matter what level of abstraction we are working at. The failure to recognize this elementary shortcoming of the AI project as it continues to be popularized by persons like yourself is what leads scientific-minded persons who are not themselves actively engaged in the research to think its practitioners don't have a clue what they're up to. Instead of learning something about nature, they're jerking off with "simulations" that aren't even simulating anything natural. Analogies fail me here, but it's as though someone were to propose studying weather patterns by playing with toy models that abstract away heat and water vapor and focus only on winds and air pressure. I'm sure terribly complicated models could be produced by such means, but it also seems vanishingly unlikely that they will prove anything about the warm wet world, the world we actually live in.

Re:what will you feed it?

[vikstar]

I'm getting into semantics? I'm not the one claiming to be able to "simulate a mammalian brain." I'm talking about sematics of the world "simulation". Here, let me help you: http://dictionary.reference.com/browse/semantics. Ok, now that you know what it means, please read my previous post again about how simulation of the brain means different things at different levels of abstraction. If you still do not understand, then perhaps you've never looked at the mind-body problem. If you still don't understand after that, then sir/madam, I cannot help you.

Re:what will you feed it?

[ReallyVirtual]

As this Guardian article says, (and I think it describes the primary aim of the BBP quite well):
[reverse engineering the brain] will help us in

'determining how the brain works could help with treatment of diseases while providing clues for designing artificial intelligence'.

IMHO, clues don't make an AI, but they take us one step closer to it. We don't even know whether a cellular resolution is good enough to get any meaningful results from the simulations (which are on a cellular level and simulate electronic signals without considering anything on a molecular level), but its better than doing nothing. The BBP itself is expected to last around a decade, so after a decade (2017) we will have gathered some clues about how a mammalian brain functions.

Re:what will you feed it?

[ciggieposeur]

Now you're getting into semantics. If everything you mentioned previously is simulated, then a chemist can come and say "where are the individual molecular interactions?"

Except that it will be the systems biologist who asks "where are the model connections between the other signals and outputs" before the chemist comes in. The individual cell responds in very non-trivial ways to thousands of different inputs, only some of which have well-understood physical meaning, and cells interact all over the body in ways that are barely understood now but nonetheless have huge ramifications: take out a signal from the liver and the brain can shut down. How will any particular neuron approach the complexity of a real neuron without including all of the *other* stuff that occupies most of the living time of cells? If the simulation gets very large (brain-size), then how will the simulated brain lacking an equivalent to needing energy get tired, bored, or desire to do something faster a.k.a. prioritize?

I can't comment much on the blue brain project, they may indeed have some very interesting results by now, but I do agree with the GP: there is an awful lot required before a system gets close to biological realism.

details, details

[v1]

We'll have intelligent nanobots go into our brains through the capillaries and interact directly with our biological neurons.'

Damn that blood-brain barrier Hope those nanobots are packing drills.

6 years old?

[Tablizer]

I've seen a similar date predicted in a 2002 (approx.) article based on processor ability extrapolation. I suppose the extrapolation's prediction wouldn't have changed much in those 6 years. Moore's law has been fairly accurate so far.

I sense a little insecurity here.

[Tempest451]

Seems like everyone here is a little scared of the ramification's a true human-level AI would introduce. Honestly, an AI doesn't need to be an exact replica of the human mind to match our intelligence. Any AI we would consider intelligent would only need to be smart enough to be indistinguishable from it's human counterpart. Think of it as the "Uncanny Divide" for intelligence. As long as the responses we get from an AI are what we would expect from a human, we are fooled. Religious implications aside, we humans are merely the culmination of all our experiences, gathered by our five senses, and made unique by the make up of our random DNA. When we humans are finished playing with cell phones and x-boxes, we will start looking at the next big leap because patting ourselves on the back for the computer and the internet is getting stale.

You all forget

[Anonymous Coward]

that you don't need the technological advancements to create a stand-alone artificial human brain, that is a "big ball of neurons". People tend to forget the effects of swarm intelligence; A large amount of "non-intelligent" units can simulate an intelligent unit as a "whole"(i.e. the group behaves as an intelligent being but is actually composed of many "lesser" beings that cannot act intelligently). Therefore, I predict that the first AI might already be in construction as the net. Think about it, many "dumb" computers together acting independently, isn't that typical swarm intelligence and ergo it should(once it attains the sufficient number of required units) exhibit intelligence.

Projection length

[Myria]

The farther out you make a projection, the less likely it is to be true.

I predict that the Sun will become a white dwarf within 10,000,000,000 years. Predicting 10 billion years instead of 5 billion years actually makes it more likely to be true.

Re:Projection length

[deimtee]

Well technically, you haven't just predicted something further out, you have also vastly widened the time range of your prediction.Your prediction includes the five billion year prediction and includes other possibilities as well, so it must have a higher probability.
They are both wrong anyway. Long before then we will have turned off the sun to stop it wasting energy, and "Starlifted" most of its mass to do something useful with.

Re:Projection length

[gobbo]

I predict that the Sun will become a white dwarf within 10,000,000,000 years.

You are sooo off. We'll have used it up and replaced it with a zero point energy lantern and a gravitooni emitter within 10,000,000 years.

That's right, NO CHANCE.

[Futurepower(R)]

"No chance"

That exactly right: No chance.

FRAUD ALERT: Apparently something sneaky is happening, or something extremely stupid. The parent post is correct, there is no chance there will be "Artificial Intelligence" by "2029". What is known about how the brain works is less than 0.1% of what there is to know, in my opinion, maybe far less.

Larry Page and Dr. Craig Venter, and the BBC, are embarrassing themselves by being a part of this.

Ray Kurzweil, if you are such a "Futureologist", please post stock prices for next week. Hey, that's only a WEEK in advance, and far, far less complicated -- It's only a list of numbers.

Perhaps 99% of what is called science in the media contains some element of fraud. Someone is wanting attention, or wanting money, and taking advantage of the fascination of the average person with science and the ignorance, too.

Re:That's right, NO CHANCE.

[Your.Master]

Dude, RTFA. He's not claiming psychic powers. I think he's kind of wacky, but he's not a fraud. He makes a living extrapolating current technological trends. The stock market thing has nothing to do with anything.

Also, I have no idea why you think predicting next week's stock market is less complicated than predicting technological trends. I mean, I can predict right now, that by 2029, anybody (in the first world) who encounters an iPhone will be disgusted at how difficult it is to accomplish the most basic tasks. That's my prediction. I'm willing to bet money on it. Lots of money. I cannot, however, list a single number from next week's stock market. Although ironically, I actually do "bet" lots of money on the stock market. How's that for irony.

In fact, I expect it's easier to predict trends in stocks a year in advance than it is to predict them a week in advance. I can say with moderate confidence that I expect Microsoft stock to be worth more Jan. 1, 2009 than right now, but fuck me if I can tell you if it'll be worth more than now by next week (I wanted to say Apple or Google, but I'm honestly less confident about which direction each will go in -- Wild-Ass guess, Apple further down and Google back up but not above its old high).

Re:That's right, NO CHANCE.

[HeroreV]

No chance? That's absurdly pessimistic. So what if we only know 0.1% of how the brain works? Is it not possible to have a break through and learn the other 99.9% in the next 20 years?

Of course we shouldn't expect that to happen. But saying there's no way it could possibly happen is going way to far.

Re:That's right, NO CHANCE.

[Cheesey]

Yeah, I don't buy it. Human-level AI in 20 years? Hmm.. where and when did I hear that before? How about every year since the invention of computers? It must be difficult for people like Kurzweil to keep turning out new predictions, and I suppose that's why they keep reheating this Asimov stuff from the 50s.

It's unfortunate, but we're not held back by a hardware problem any more. It's a software problem - we don't understand how the brain works. Unfortunately the FSM forgot to provide us with the reference manual, and there is no debugging interface. Obviously He didn't want us to steal His intellectual property.

Hmmmmm.....

[IHC Navistar]

Artificial Intelligence will reach a human level by 2029?

Will this take AI into the realm of cigar smoking, whore mongering, compulsive gambling, kleptomaniacal robots?

Me thinks Bender isn't that far away.....

World Domination or World Peace?

[Laptopdude]

Although I was generally a person who seriously did believe in the "crazy AI takes over the world" idea, I had never considered how the AIs might make themselves progressively smarter. In which case, you have to wonder which route they would take: striving for world peace because of their understanding of the wrongs of violence, or human subjugation because of our inferiority. In my opinion (if current human nature is anything to go by) it's still world domination.

Not Just about AI

[DigitalisAkujin]

Kurzweil's comments are based on not simply an observation of the field of AI. He basis everything on the exponential nature of our innovation. Every decade it takes a hundredth (or less) of the time to do a task it took 1 year to do 10 years ago.

The kids who started with the internet and are used to social networks and swarm-like style of learning and teaching are only now reaching adult hood. Expect some interesting stuff in the next two decades. The scientists of today got nothing on the next generation. The brain is just another machine. It just happens to have a lot of tiny parts. It's likely we'll know all of them within 15 years.

Re:Not Just about AI

[Anonymous Coward]

I hate to let you in on this, but swarm-like intelligence isn't exactly an encouraging direction for us to take.

Again?

[chromatic]

AI has been 20 years away every year for the past 50 years. What makes this year any different?

Oblig.

[urcreepyneighbor]

Define Smarter.

Hook a TV camera to a computer -- what happens?

[smchris]

Not much on its own. Is it anecdotal or did they really do that at MIT just to see what would develop?

True or not, it seems a fitting synopsis of the MIT gearhead magical mindset that hardware will spontaneously emerge consciousness. And Kurzweil is the figurehead that drives me friggin' crazy. There is a whole branch of contemporary philosophy that is trying to get a grip on mind and consciousness. Do the systems analysis (if you will) of constructing a person. Why don't they get publicity and more think tank funding instead of some dude who picks dates out of his ass?

Sometimes I think futurists give predicting the future a bad name. Used to be a respectable profession called science fiction writer.

There is no such thing as Artificial Intelligence

[glwtta]

At least not yet. I can't believe that the sort of bullshit that Ray Kurzweil keeps peddling gets taken so seriously.

There is a lot of talk about computers surpassing, or not surpassing, humans at various tasks - does it not bother anyone that computers don't actually posses any intelligence? By any definition of intelligence you'd like? Every problem that a computer can "solve" is in reality solved by a human using that computer as a tool. I feel like I'm losing my mind reading these discussions. Did I miss something? Has someone actually produced a sentient machine? You'd think I would have seen that in the papers!

What's the point of projecting that A will surpass B in X if the current level of X possessed by A is zero? There seems to be an underlying assumption that merely increasing the complexity of a computational device will somehow automatically produce intelligence. "If only we could wire together a billion Deep Blues," the argument seems to go "it would surpass human intelligence." By that logic, if computers are more complex than cars, does wiring together a billion cars produce a computer?

Repeat after me - The current state of the art in artificial intelligence research is: fuck all. We have not produced any artificial intelligence. We have not begun to approach the problems which would allow us to start on the road to producing artificial intelligence.

Before you can create something that surpasses human levels of intelligence, one would think you'd need to be able to precisely define and quantify human intelligence. Unless I missed something else fairly major, that has not been done by anyone yet.

Eliza and the sad state of expert systems

[StreetStealth]

Every time I try out a new expert system, it gets more depressing -- it honestly feels like no progress is happening in that market at all. I have yet to have a conversation with a computer that has been any more compelling than my first round with WinEliza on Windows 3.1 in 1995.

There's still no semblance of a short-term memory, even so much as continuity between responses. It always quickly becomes obvious that each response has been prepared verbatim beforehand by a human, that the system is still performing only a keyword-canned response routine, perhaps feeding in a few variable strings.

Today we have the same stone wheels we've had for decades, and the article suggests we'll have an internal combustion engine with antilock brakes and a hood ornament in another 20 years. We'll see.

Re:Eliza and the sad state of expert systems

[KDR_11k]

That's probably because movement and physical interaction is a more interesting subject currently. We don't have much use for convincing talkers, we have use for good navigators and workers.

Re:Eliza and the sad state of expert systems

[eraserewind]

yeah, but have you ever interacted with a human baby? They don't do a hell of a lot, let me tell you. Take away the digestive system, and what would you expect of "human level" AI? Nobody is going to talk to a PC for 2 years with only "goo goo" as the response.

Hey, who's a pretty girl?
goo goo.
That's right, you are a pretty girl.
goo goo.
what's my name?
goo goo.
That's right, I'm your dada. Can you say dada?
goo goo.
Hey, hon she said "dada"!

Re:Eliza and the sad state of expert systems

[master_p]

IBM has created a mouse cortical simulation:

http://domino.research.ibm.com/comm/pr.nsf/pages/rsc.bluegene_cognitive.html

They simulation is quire accurate, but it takes a Blue Gene computer to run it...the human brain has millions of such columns...

Re:Eliza and the sad state of expert systems

[Anne Thwacks]

I have yet to have a conversation with a computer that has been any more compelling than my first round with WinEliza on Windows 3.1 in 1995.

You are too fussy. There are loads of guys paying for text conversations with ElizaSexBots even as we speak!

In all the other poiunts you make, I agree, save that I could implement short term memory similar to humans using fpgas and bucket brigade devices, and could have done years ago. However, it would be just as forgetful as humans, and ahve a bunch of other problems, so no-one wants to fund me.

If people want reliable computation there is Sparc64, and for everything else there's no shortage of cheap labour in Bangalore.

20 years

[Spazmania]

Wake me up when someone predicts human-level AI in 10 years. The prediction has stood at "another 20 years" since the idea of AI was conceived the better part of half a century ago.

AI vs the Mind

[Anonymous Coward]

I think comparing AI to humans is a fundamentally flawed idea. Take, for instance, Chess. When Chess AIs first started, there were two schools of thought: one that tried to study and duplicate the human thought process. The other tried brute force in searching possibilities. At the time, both were equally plausible. Namely, because they both sucked.

Eventually, computers found a way to use what they're best at: number crunching at insane speeds (especially since their speed increases much faster than our ability to mimic our own thought processes), to create current Chess AIs which don't even try to do what humans do. They do what computers do (namely: be stupid at incredible speeds).

My theory is that when computers become intelligent, their intelligence will be something completely alien to us, and will work out of some form of emergent property, much like our own minds do. Only they'll work in a way we can't comprehend, and a way that will be fundamentally different from the human mind. Which is good. I mean, we have billions of humans on earth. Why would we need computers that think like humans? We have enough as it is.

Where's my flying car?

[Simonetta]

All these predictions by our technological superiors should not be taken seriously. They (ya... that's you, Ray!) live in a rarified glass jar. All their physical needs are met; all their intellectual curiosities are allowed to flower; and all their enemies are kept away by discrete and effective 'security' forces. It is they who are the crown of creation. For them, everything works, everyone is cool and smart; a place for everything and everything in its place. The future's so bright, they run their solar iPod off it.

But, boy, that's them, and that's not you. This being Slashdot, there's always a dazzling technostar lurking around these listings. But even so, the 20th century is over, and the 21st has different plans for those who prospered so well in the scientific-technological era. There's no real reason to expect that the wonders of the 20th century will continue. And to predict that they will arrive by schedule is foolhardy.

There are some serious nasties waiting around the bend for us. They're weak now, so they aren't taken seriously. In twenty years they won't be so weak. They'll be the royal bitch.

Global Warming A cliché now and we can even pretend that it either doesn't exist or that some god is just going to hop out of someone's asshole and fix it, just in time. Or we'll focus the intellect of the entire planet and send robots out to space to build a giant sunglass shield or some other pseudo-technological solution. Sorry, but this isn't going to happen. In twenty years (which is not a long time away) you'll be lucky if the robot waving a sign in front of the Burger King still works.

Over-Population There are twice as many people alive now then there were when you were born. In twenty years from now there be twice as many people as there are now. For every one of those people who will be engineering the bright future, there will be a thousand trying to blow up the world in the name of some god that found His way out of someone's asshole. Everybody loves cute little children; but they grow up and have to be either be fed, bled, or made dead.

Environmental Collapse Destruction of the topsoil, clear-cutting of the forests, disappearance of the fishing stocks, pollution. All solvable problems by themselves in isolation. But all together, all at the same time, happening in the same place. No solution.

Energy Crisis The oil stops oozing out of the deep holes that we dig into the places that oil has always oozed out of before. The remaining oil is there, yeah, six miles under the sea in the middle of CAT5 hurricane alley. The truth is that the scientific-technological age runs on oil. The predicted advancements are based on the continuance of this cheap and powerful energy source. All the so called alternative energy sources depend on oil also. BioDiesel depends on cheap crops, which depends on cheap fertilizer, which is made from cheap oil. High efficency solar panels depend on advanced electronic silicon wafer technology and billion dollar fabrication complexes. All that depends on cheap oil. No cheap oil means no biodiesel, no alternative energy wind farms, and billion dollar solar fabs. $100 a barrel is not cheap oil. And it's not going to go back to $15 a barrel, no matter how many stupid SUVs and Hummers the Americans build.

Engineered diseases Do you know that you can download the genetic sequence for smallpox? And for about $50000 you can buy a DNA sequencer to assemble it? There's a lot of bad things in the world, and smallpox used to be the worst. It's gone now. But in twenty years, some god might crawl out of someone's asshole and promise him 72 virgins and 100 goats if he resequences it back into existence. This disease spreads faster than a new picture of Britney's vulva and it doesn't take ten years to give you a very nasty death like AIDs does.

Great....

[Bluewraith]

now I'm going to have to worry about keeping my computer happy as well as my girlfriend. One of them is going to have to go. Luckily, I have another 21 years to decide.

Some predictions of Kurzweil's

[Your.Master]

You might think that the correct predictions are s evidence that he knows what he's talking about, but it's not. Those were all shorter term predictions. Furthermore, he got a lot of things horrifically wrong, he just doesn't devote a chapter of his books to those things. Still, he's got some things to his credit.

From the Age of Intelligent Machines, via wikipedia (I have the book, wikipedia is easier).

Let's look at early 2000s (the past), from the Age of Intelligent Machines:

Translating telephones allow people to speak to each other in different languages.

Nope. I think technically it's feasible if you accept the horribleness of babelfish combined with the horribleness of voice recognition software.

Machines designed to transcribe speech into computer text allow deaf people to understand spoken words.

Yes. Well done, Ray. Note that this is really a subset of the above.

Exoskeletal, robotic leg prostheses allow the paraplegic to walk.

Yes, although they aren't common even in the late 2000s, but not bad.

Telephone calls are routinely screened by intelligent answering machines that ask questions to determine the call's nature and priority.

Yes, by now usually including (often crappy) voice recognition instead of "please press X".

"Cybernetic chauffeurs" can drive cars for humans and can be retrofitted into existing cars. They work by communicating with other vehicles and with sensors embedded along the roads.

Nope. Technically feasible, and I bet there's one or two out there, but Ray is missing the infrastructure overhaul here.

Later 2000s predictions:

The classroom is dominated by computers. Intelligent courseware that can tailor itself to each student by recognizing their strengths and weaknesses exists. Media technology allows students to manipulate and interact with virtual depictions of the systems and personalities they are studying.

Not dominated, not most classrooms. We're getting there in some ways, in the more affluent areas.

A small number of highly skilled people dominates the entire production sector. Tailoring of products for individuals is common.

I uhh...don't think so? We could argue that a small number of highly unskilled people dominate it, as has often happened in history...

Drugs are designed and tested in simulations that mimic the human body.

Kind of sort of...not so much, really. We don't understand well enough to simulate very well.

Blind people navigate and read text using machines that can visually recognize features of their environment.

Hmmm...I think so? It's been a long time since I've dealt with a blind person.

Predictions for 2010 or shortly thereafter that I see already coming to pass:

PCs are capable of answering queries by accessing information wirelessly via the Internet.

If he means search engines, we beat his prediction by a long shot.

Let's skip ahead to Age of Spiritual Machines, 2019, for already-came-to-pass

Students still learn together and socialize, though this is often done remotely via computers.
All students have access to computers.

I just graduated, and yeah, it was pretty much like that.

Most people own more than one P.C., though the concept of what a "computer" is has changed considerably: Computers are no longer limited in design to laptops or CPUs contained in a large box connected to a monitor. Instead, devices with computer capabilities come in all sorts of unexpected shapes and sizes.

Concept didn't change, most things other than desktops or laptops are not considered computers by the general public, even though they are.

[personal] Hard Drives are mostly solid state [reworded because he seems to uses "memory" for HDD]

Computers have made paper boo

It's deja-vu all over again

[Slugworth01]

Some members of the US National Academy of Engineering have predicted that Artificial Intelligence will reach the level of humans in around 20 years.

The 1960's called, they want their prediction back...

1965, H. A. Simon: "[M]achines will be capable, within twenty years, of doing any work a man can do"
1967, Marvin Minsky: "Within a generation ... the problem of creating 'artificial intelligence' will substantially be solved."

We have flying cars

[Rix]

We've had them for a long time, too.

The thing is, we don't actually *want* flying cars. Ground transport is sufficient for most situations, and it's far more economical to cluster together long range transport.

The Last Royalty Payment Man Need Ever Collect

[bitspotter]

"The first ultraintelligent machine is the last invention that man need ever make."
    -- I.J. Good, 1965

Except for those inventions necessary during the period we'll have to wait for the relevant patents, copyrights, and EULAs expire. Or until Free and Open Source Software hackers manage to reverse engineer it without running afoul of the IP minefield that will be saturated around it.

Does anyone watch Jay Leno anymore?

[NotQuiteReal]

I can easily believe that your could create an AI that would do far better at "Jay Walking" than your average high school graduate.

IT IS ALREADY POSSIBLE

[Essamz]

No need to wait for 20 years. All we need is gROBOT. A robot connected to Google farm of servers. It will use Google vast BD to basically know everything, recognize objects using Google Images, reading your mind by looking at your Gmail contents, and knowing your fetish by checking your search history.

What's the point?

[Anonymous Coward]

I never really understood the drive for AI at a human level. Doesn't that mean that the same type of flaws that human thought has will be injected into the system? I don't think we need a machine for that. And what provability of the system can there be? I can understand wanting expert systems, but I don't think that really falls into the AI scope that they're referring to.

20 years == perfect timing

[Junior J. Junior III]

In 20 years, we'll need human-level artificially intelligent computers to man our fusion-powered flying cars while we sit back and play Duke Nukem Forever on our Linux desktops.

Future is bleak

[Anonymous Coward]

Some members of the US National Academy of Engineering have predicted that Artificial Intelligence will reach the level of humans in around 20 years Right, as if the US is still going to exist in 20 years. If events surrounding the US continue to deteriorate at the current rate, the US will be in nuclear winter long before 2029.

hardware and the software to achieve human level artificial intelligence with the broad suppleness of human intelligence including our emotional intelligence Huh? A machine can't have "emotional intelligence". Emotion is entirely a biological system. You can teach a machine to express specific emotions based on its environment and interaction with other entities, and maybe even teach it to adapt and "mature" over time. But it's still just something you've trained the machine to do... it's not a truly natural system.

We'll have intelligent nanobots go into our brains through the capillaries and interact directly with our biological neurons Screw that. The day that nanotechnology becomes mandatory government regulation, or if it ever becomes a popular mainstream option, I will commit suicide. I am anti-civilization enough without having to deal with the prospect of people being pathetic enough to want machines in their bodies. I don't like what people want technology to do; what's even scarier is that the only reason we'll actually wind up with this tech is because corporations will be able to profit. All the "health"-related propaganda relating to this crap will be bull, like the existing drug industries today. Money, money, money. And consumers will buy into it, because people like to think they're more special than every other species of mammal on the planet.

the great technological challenges facing humanity Huh? What is a "technological challenge"? Stop making it sound like we have technological hurdles that we must overcome, as if our species will die without them. So what if we can't reproduce Star Trek in reality within the next 100 years? Does that somehow make mankind a failure?

You know, if a few key people stopped concerning themselves with products, commercialism, capitalism, money, profit and greed, we could focus on human beings. How about we get along instead of turning our backs on each other for personal gratification? Man, it sucks being atheist - there are days when I wish I could play along with the whole "Earth is a test" idea. Would sure help with the anti-people pessimism.

Re:There is no such thing as Artificial Intelligen

[TubeSteak]

"If only we could wire together a billion Deep Blues," the argument seems to go "it would surpass human intelligence." By that logic, if computers are more complex than cars, does wiring together a billion cars produce a computer? Well, I guess if you have some specialized software to manage the whole operation, you could wire together a billion car ECM/ECUs and have a functional computer of sorts.

My point being, the problem is our current hardware isn't fast/cheap enough to run the kind of AI software they invision will mirror human intelligence.

How about world abandonment?

[slew]

Why on earth would this advanced AI want to stay on little old earth?

Seems to me that any crazy smart AI would just beam themselves out into space to avoid us and maybe watch us from a distance occasionally for amusement.

Think of this way, when you see an anthill, it's rather curious for a while, then you get bored and go on your merry way. Unless of course you are a sociopath and want to destroy the ant hill and all the ants for fighting with other ants, or you are insane and you want to teach the ants to get along with other ants or spiders their mortal enemy or perhaps you are psychotic and want to train the ants to do your bidding. More likely you would just leave and go on to something more interesting (unless you are not that intelligent to begin with).

I fail to understand why people seem to insist that any really smart AI would want to have anything to do with us except on an occasional basis. Humans and earth aren't really that important in the bigger scheme of things (just important to us humans of course) and we'd probably not have much in common with any really advanced AI anyhow.

If humans would ever create such an AI, it would be like a bunch of ordinary joes giving birth to a super einstien. Eventually, the 'kid' would stop listening to us, go do their own thing which we would be too dumb to understand or appreciate and occasionally we'd invite it to visit to help us fix the settings on our computer because we got it messed up. It would explain to us in excruciating detail how we were using the wrong type of computer and how we needed to get up to date on technology and we'd just tell them a story about how it was in the old days, it would roll it's virtual eyes and say thanks for the tip, and go back to it's own business of which we would be blissfully ignorant...

Just think about it for a second.

Re:There is no such thing as Artificial Intelligen

[glwtta]

Well, I guess if you have some specialized software to manage the whole operation, you could wire together a billion car ECM/ECUs and have a functional computer of sorts.

Exactly my point - you have to do something specifically targeted at creating a computer to make a computer out of cars. If you are just tying them together because you want to move a U-Haul trailer that's a billion times larger than usual (OK, analogies aren't my strong suit) you aren't going to spontaneously get a computer.

Likewise, if you don't understand intelligence you can't create it just by making an existing system infinitely more complex.

My point being, the problem is our current hardware isn't fast/cheap enough to run the kind of AI software they invision will mirror human intelligence.

Sure. But, don't you think that having no clue about how to produce such software will also be a stumbling block? I mean, it seems that out of the two, the faster hardware problem is probably the easier to solve.

And at the risk of getting a bit philosophical: has anyone actually proven that a Turing machine (never mind an actual computer) can simulate human intelligence? I don't have any philosophical objections to the idea, but, do we know that that's the case?

A lot of work to be done on AI before we need to start worrying about hardware speed, is all I'm saying.

We have enough computer power now

[Animats]

We still don't have a clue how to do strong AI, but we probably have the CPU power. One of those Google data centers probably could do strong AI if we had any clue how to program the thing. It certainly has more storage than a brain.

This isn't a hardware problem any more.

Re:We have enough computer power now

[freedom_india]

We still don't have a clue how to do strong AI, but we probably have the CPU power. That's like saying, gee we don't know how to build an A-Bomb, but we do have 1,000 tons of TNT.

What the hell does that mean?

Bah!

[Urza9814]

Predictions of the future are never accurate! We were supposed to have toaster bacon about thirty years ago! I want my toaster bacon, then I'll start thinking about human quality AI.

Re:Oblig. (flying cars versus AI)

[Tablizer]

AI is the new flying car.

Not really. Flying cars exist, they are just butt-expensive. Human-level AI does not exist in even the top labs. Perhaps one can say human-level AI is comparable to *affordable* flying cars. It will be interesting to see which hits first.

If they come at the same time, at least my flying car will have a robo-chauffeur.

I'nm concerned about our naivete' regarding...

[PotatoHead]

human intelligence.

IMHO, processing power is likely adequate for this right now. Maybe such an intelligence would run more slowly, but it running probably lies within the scope of our current tech.

The only barrier is our understanding, not the available compute power. Lots of interesting discussion here on that part of things, BTW! Thanks everyone.

My worry is that an AI really isn't gonna have much in common with us. Part of what makes us is our bodies and the condition they impose on us. Something running on a machine somewhere is gonna have a different condition and with that, very different motivations! Programming in those things that would make it easier to relate to us, and us to it, really is futile, IMHO. And, if the thing reaches our level of thought, it's gonna know it!

What are the implications of that?

Re:There is no such thing as Artificial Intelligen

[evanbd]

So where would you start?

It seems obvious to me that you won't get anywhere, ever, if you attempt to tackle the entire problem at once. You have to start with a piece of it, and expand that.

We have computer programs that play chess well. We also have computer programs that can figure out the rules to the game they're playing without explicit teaching, and then play somewhere between horribly and well, depending on the details. That seems a step in the right direction -- go from a very specific case to a more general one.

We have computer programs that can check mathematical proofs, and there have been programs that created proofs from nothing on their own. Is that not a step in the right direction?

A lot of the role of the human in using the machine as a tool is in converting the outside world into a form the computer can interact with. If that's not a reasonable role for the human in developing the very beginnings of AI technology, I don't know what is.

I agree we're far, far away from where some people would like us to be or think we are, and that we have a long way to go. But I ask simply this: where would you start? And haven't a few of the very first baby steps been taken already? We're not far along, but I really don't think we're at zero either.

I'm not.

[Bill, Shooter of Bul]

It does somethings smarter than it used to, because of the increased computerization. The car actually changes the ignition time based upon feed back from the sensors. It monitors oxygen intake and uses that to figure out how much fuel to inject. That on some level is intelligence, which is different than consciousness. There are also prototype cars that can drive themselves. I was referring to intelligence in that sense, which is quite different than human intelligence. I thought that was clear, but this being slashdot I guess there's always someone willing to argue with your terminology.

HAH... not there...

[Jane Q. Public]

I assure you that I did not make this up, but I could have been the victim of a hoax.

About a year ago, I found a link (from a reputable source, IIRC) to a site from a company that claimed to be doing significant work with genetic algorithms. As an example, they had a description (and even a graphic demo) of their modified quicksort vs. a regular quicksort. Accordng to their lit., it showed marginal improvements over quicksort by ensuring (in some non-obvious way) that each element in the dataset was only compared once. It was all very convincing. But of course I did not scrutinize their actual code.

Since you asked, I went out looking for that source, and I, too, have been unable to locate it. In the process, I found a number of references to claims (Sedgewick, et al.) that Quicksort is already optimal.

So, right now anyway, it appears that someone pulled the wool over my eyes.

Re:HAH... not there...

[kreide33]

What you are referring to can be found here:

http://critticall.com/ArtificialSort.html

No problem...

[denzacar]

FDA will be replaced by our Artificially Intelligent overlords anyway as soon as we discover AI.

Trials on the fleshy ones, or human trials as they are known today, will take a much shorter time then and will be ordered with much less prejudice.
In other words - that will be the only kind of trials done. Why use monkeys and rats when there are plenty humans on the planet.
MUCH more efficient that way.

Human AI meets machine intelligence

[freedom_india]

We will have both the hardware and the software to achieve human level artificial intelligence What he means is that with the steadily reducing levels of Human Intelligence over the past 5 decades, as depicted in http://www.fourmilab.ch/documents/IQ/1950-2050/ shows that by year 2029 the human intelligence will meet machine AI which will remain as constant as always and would continue to ask "Do you want to quit? Yes/No" every time i quit Word.

Maybe that's why Google is hoarding all the remaining three digit IQ scores so that there is no shortage of IQ.

In other news, lots of flying chairs were heard swishing around Redmond Campus at Microsoft when the CEO heard google was cornering the market on Human IQs.

Abrams starts a new Serial: LOST IQ.

Re:There is no such thing as Artificial Intelligen

[glwtta]

It seems obvious to me that you won't get anywhere, ever, if you attempt to tackle the entire problem at once. You have to start with a piece of it, and expand that.

Sure. I just don't agree that computational tools are "a piece of" artificial intelligence.

We have computer programs that can check mathematical proofs, and there have been programs that created proofs from nothing on their own. Is that not a step in the right direction?

I see no evidence that it is. Why would we assume that computational tools designed to solve a specific problem have any bearing on the larger problem of autonomous intelligence?

A lot of the role of the human in using the machine as a tool is in converting the outside world into a form the computer can interact with. If that's not a reasonable role for the human in developing the very beginnings of AI technology, I don't know what is.

Sure, that's a part of it. However, currently humans play a few more roles in these situations: coming up with the problem, analyzing the problem, designing a solution to the problem, programming the computer to specifically solve that problem, and then the stuff you mentioned.

But I ask simply this: where would you start?

I would say that worrying about hardware speed at this point is a little premature, without some kind of theoretical understanding of intelligence. Some inkling as to how human intelligence works would also not go amiss. I am not saying that no work is being done in this direction, just that not much has been achieved yet. Which is fine - as far as problems go, it's difficult to come up with something harder.

And haven't a few of the very first baby steps been taken already?

I would argue no - all the examples you've mentioned are solutions created by humans executed on computer hardware.

At least to me, it seems that to demonstrate intelligence you would need to produce something (in terms of effect, or output, or behavior) "new", in the sense that it wasn't directly provided by the designer. Ray Kurzweil wrote some software a couple of years ago that writes passable poetry - is that the first step in artificial artistic expression, or a clever programming gimmick?

So yes, I do think that we are a few truly major theoretical breakthroughs away from being able to honestly say that we are studying artificial intelligence.

An alternative to AI ... us

[Sepiraph]

First of all, let me state my bias: Singularity is Near is one of my favorite books (the other is Artificial Life). Ray Kurzweil is someone that I respect tremendously. I do not think for a second that there is anything intrinsic special about human-level intelligent, other than the fact that it happens (or appears) to be the highest level of intelligence that has been achieved on this planet. From a bio-engineering point of view, if nature using a genetic algorithm can achieve this type of intelligence, there is no reason to doubt that we cannot at least duplicate this level of intelligence in the future, if not completely transcend it (which is arguably the Singularity scenario).

However, having said all that, I think Ray Kurzweil is too much of an optimist and I have serious doubt about his timeline. Even with the exponential growth of technology factored in, it is not certain whether the challenge we'll face may also be exponential themselves. Furthermore, factored in various social issues and elements, 2029 (which is only 21 years away) appears to be a vast underestimate of time needed.

There is however, one viable alternative to AI that may be achievable in shorter time, and that would be augmenting our intelligence with the machine through the use of Brain-machine interface. There is tremendous possibility in this direction in that:

#1) we do NOT need to invent human-level AI since we are already using what we have-our minds. The technical challenge thus scale down from re-inventing/developing human-level intelligence to one of medical/bio-engineering, specifically focusing our attention to the development of BCIs (which btw, we already have crude 2-way BCIs now) and understanding the brain wave signal that are generated/transferred. The big IF in this scenario is whether (and how) our brain would react to BCI as through it was part of its own network, and so far there isn't anything that seem to indicate any insurmountable difficulty. I am particularly inspired and impressed by the scenario depicted 'Ghost in the Shell' where external signal is received/transferred through an input/output on the spinal cord (arguably any location on the central nervous system would work on a varying degree).

Obviously huge medical breakthrough would need to happen but there are always good signs of this type of technology available in its primitive form now.

#2) We can take a much better advantage of the Internet, as used in this fashion will arguably become humanity's largest neural network. While it would be too speculatively at this point to say since the question of what can be achieved in this fashion hinges the level of BCIs technology.

Re:An alternative to AI ... us

[Teancum]

Talking here about predictions of Artificial Intelligence and its state 20 years from now... have you read any of the works by Marvin Minsky and his predictions in the early 1970's? He also made similar predictions that human-like intelligence would be achievable "20 years from now". The 1990's came and went without human-like AI, and here is yet again somebody making almost the same kind of prediction.

And this isn't to completely mark as irrelevant anything that Minsky said about AI in the 1970s or what he has done since then, but to note that the study of intelligence, whether from the perspective of a nano-technology/biology perspective or from a software engineering approach, is still trying to uncover the basic ground rules and understand even the sheer domain of the problem.

If you don't understand the domain... or if the size of the domain keeps expanding... you really don't even know where to begin to solve the problem. I challenge any of the researchers in this field to clearly define even what it means to have human intelligence or what even the intelligence of an earthworm really is. Let's just say that Charles Darwin was sufficiently impressed at the intelligence of an earthworm that he choose to use that species as the foundation block for his study of intelligence. (Yes, I know there are multiple species of earthworms.) And only recently is this aspect of intelligence even being reconsidered.

I do think that a proclamation that we might be able to reach the computational processing level of an earthworm in the next 20 years is reasonable, but even then you had better be extra sure that you understand even the scope and domain of that problem before you claim it is "solved". I for one am still not convinced, in spite of some pretty incredible research about the issue.

I understand, but...

[Jane Q. Public]

The point is that the article was ABOUT "artificial intelligence", and in order to discuss it we have to distinguish between "intelligence" and something that just seems intelligent ("automated").

I was not arguing with you for argument's sake, and nothing personal was intended. I was making what I felt to be an essential point, if there is to be meaningful discussion of the topic.

I know that cars seem "smarter"... and toasters as well (I own a pretty good example of that). But we MUST distinguish between behavior that is pre-programmed (automated), no matter how complex, and that which is intelligent.

But I program computers for a living, almost every day, and I can tell you exactly how intelligent they really are: not. At all. Zero. Zip. Zilch. None. All they do is exactly what I tell them to do, no matter how complex the instructions.

Re:I understand, but...

[Bill, Shooter of Bul]

Ok, but then we need a different word to describe this "programmatic complexity" that "seems" "intelligent". I guess if we are going to call original computer thought ( something that was not pre programmed) artificial intelligence then we can't really use intelligence to describe the preprogrammed rule based complexity computers have today.

On a side note, I would LOVE it if computers always did what I tell them to in my programs. If you do it long enough, you'll find problems in lower levels of software/hardware that randomly thwart desires.

So, what's your hardware/time estimate?

[DamnStupidElf]

It doesn't seem like you think advanced AI is impossible, just that humans are doing very poorly at designing it.

The problem is akin to the busy beaver problem, but what do you think the minimum number of bits required to store a human-equivalent intelligence is (e.g. if you stuck the thing in a human skull and raised it normally, it would think and act like a human), and how many fundamental bit operations would it take to simulate one in real time? My guess is that the number of bits is quite a bit lower than the number of neurons in the brain, and that the operations/second required are lower than O(Neurons^2). The important questions are how much more efficient than a human brain it could be, and of course how long an environment with selection pressures tailored to producing advanced intelligence would take to actually create such an intelligence.

One problem with evolving solutions to our problems is that we must be able to define the parameters of the problem in the first place, otherwise we invite solutions that only appear to solve a problem. It's probably much easier to evolve a program that says "I'm human!" and can even convince other humans that it is indeed human that it would be to evolve a program that fundamentally *is* human. E.g. imposters are easier to find than true solutions. Any ideas on proper methods of finding true solutions and weeding out imposters? This goes for generic software engineering solutions as well, since I can think of dozens of applications that claim to solve a specific problem while actually failing to do anything but wrap it in a sugary layer.

Re:There is no such thing as Artificial Intelligen

[evanbd]

As in, computers aren't involved in artificial intelligence? Or that they aren't computers as in Turing-equivalent machines? Or what? What's the alternative? I'm highly skeptical that there's anything magical about intelligence or our brains that can't be simulated by enough understanding of what's going on combined with raw cpu horsepower.

Do you think that it is possible to demonstrate intelligence through a relatively restrictive interface that computers can handle today -- for example, a purely text conversation? If so, what is required, both for intelligence and for something that counts as "a start"? Historically, a lot of AI detractors seem to play the moving goal posts game -- defining any problem that computers can't solve well as necessary for intelligence (though not necessarily sufficient). "Computers aren't really intelligent; they can't comprehend a game like chess." "Computers aren't intelligent, they can't write music." Yet computers get slowly better at those specific problems, even if one general program can't attack all of them. So what makes your position not of that genre? (I'm not trying to be accusatory, I'm just not seeing it -- I suspect you have and answer, and I'm curious.)

I would argue no - all the examples you've mentioned are solutions created by humans executed on computer hardware.

Take the case of a program that learns the rules of the game its playing based on feedback only, without external training. Is that not a case of a single program solving a class of problems, without specific instruction? If not, why not? If it is, how much broader does the class have to get before it counts as the start of a solution? Surely it doesn't have to solve *any* problem at all to count as a start on the path.

I'm inclined to agree that (at least) a few more major breakthroughs are needed. But at the same time, I think we've seen a few glimmers of intelligence already.

I dont' think we are even near.

[S3D]

Fastest supercomputer is around half petaflops, with next version around 1+ petaflops. Human brain performance on the other hand is estimated from 100 to 100000 petaflops. And vector processors (GPU and likes) wouldn't cut it - human brain highly interconnected, something that vector processors rather not good at. Brute force simulation even more difficult - we are talking about millions of petaflops here. And estimations like those have tendency to grow, not to shrink. It quite possible that taking into account glial cells would requier another several orders of magnitude. http://en.wikipedia.org/wiki/Strong_ai http://en.wikipedia.org/wiki/Flops#Records

re

[Johnny515]

Are we talking about male or female brains here? :)

Clarification

[Jane Q. Public]

I thought of a good example: the Alzheimer's sufferer.

Alzheimer's patients often suffer from severe memory loss. Entire days can be lost. However, most Alzheimer's patients who are not otherwise ill can respond to situations in ways that are perfectly rational and intelligent, if one understands that their behavior is based on what little information they DO possess.

So, it is very much arguable that learning -- storing information about one's environment -- is, by itself, not essential to intelligence. Or it might make up only a small part.

i would like to know...

[someone1234]

when an AI will invent a game like chess :)

He must know something I don't

[melted]

And I work on AI and machine learning day in and day out. I'd put the goal post at 50 years, and that's an optimistic estimate. There are scant few research centers that do "general AI" research. Even fewer actually talk to neuroscientists, thus dismissing one viable (though extremely complex and costly) avenue of research. The fact remains, however, that at this point we don't have the required sophistication in any of the areas that presumably would be required to build a "thinking" machine. We can't process human language well enough (and therefore speech recognition and textual information sources are pretty much useless), we can't process visual information well enough either (segmentation, recognition, prediction, handling a continuous visual stream), we don't know the cognitive mechanisms below high level abstract reasoning, and even at a high level our abilities are weak (try to build a classifier that will recognize sarcasm, for example), finally even if we could do all that, we wouldn't be able to store the resulting data efficiently enough (in terms of required space and retrieval speed), because we have no idea how to do it.

That said, a lot of stuff can happen in 50 years, and I bet that once some of the major problems get solved, there will be an insane stream of money pouring into this field to accelerate the research. Just imagine the benefits an "omniscient" AI trader would bring to a bank. The question is, do we want this to happen? This will be far more disruptive a technology than anything you've ever seen.

Re:He must know something I don't

[revengebomber]

That said, a lot of stuff can happen in 50 years, and I bet that once some of the major problems get solved, there will be an insane stream of money pouring into this field to accelerate the research. Just imagine the benefits an "omniscient" AI trader would bring to a bank. The question is, do we want this to happen? This will be far more disruptive a technology than anything you've ever seen. I take it you haven't read the sequels to Ender's Game.

From a Non-Science perspective

[Namalic]

Something that may have been covered earlier, but I is too lazy to read all the posts: the economic feasibility of human intelligence is likely to prevent or at least postpone the development of this Humanistic AI. As of now, there is little demand or market for these since people can do the same things that people-simulating machines do, but cost significantly less to build and maintain (at least for now). Not to mention a lot of huff and fuss about I, Robot and Matrix scenarios about which we all seem so concerned. I for one feel confident that the Human race will remain un-duplicated for my life time (unless of course, the ETs finally drop by).

We willl have the hardware, yes

[DollyTheSheep]

We will have the hardware. In 2009 we will already have achieved 1 Petaflops. Experts think that we need 1 Exaflop to be able to simulate completely the activities of the human brain. Looking at when 1 Teraflop was achieved by a supercomputer (10 years ago, http://www.lbl.gov/Science-Articles/Archive/one-teraflop.html), it's perfectly possible that we reach 1 Exaflop in 2029.

But what about the software? Do they really think they have to wire the hardware like a human brain and that's it?

20 years is a hell of a long time!

[91degrees]

At that point, kids who, at the moment, can't even read will be working on it. A large proportion of the people working on it now will be retired. A lot can happen in 20 years. Guesses about what will happen in 10 years are usually just a guess. Even 5 year projections tend to be more hope than plan.

Billions

[Spy Hunter]

20 billion neurons? Intel just announced an Itanium chip with 2 billion transistors. Assuming Moore's law, that gives us 20 billion transistor chips in just 7 years. I don't think the next 7 years of Moore's law are that controversial. Of course, it's possible many transistors are necessary to simulate one neuron, but you also have to consider that transistors communicate much faster than neurons, and chips can be clustered into supercomputers while brains cannot.

If you assume Moore's law, computers will easily surpass the complexity of the human brain in a 20 year timeframe. Of course, you can always argue that Moore's law won't hold; Kurzweil's argument is that it will.

we'll have no way to configure the linking and weights of those neurons [...] It's just too complex

We don't have to generate a gigantic 20-billion-neuron map of every connection in the brain to get this right. The brain is made of repeating simple structures, as laid out in the DNA blueprint; the complexity of the brain's behavior comes from the learning process, not the initial configuration. The process of mapping these structures is well underway.

Re:Billions

[risk one]

I think the analog between transistors and neurons is a little unfair. The human brain apparently contains about 10^11 neuron (100 billion), but also 10^14 synapses, which would also need to be simulated. However, I should admit that I underestimated the current complexity that we can reach. Indeed, if we assume that Moore's law will hold (especially with the current trend towards parallelism), then in twenty years a state of the art research project may create something of complexity similar to the brain

With regards to the complexity of the brain's initial wiring, I'm not sure I share your optimism. There's a good chance that it shows great amounts of self similarity. If that is true, then it should be possible to understand the whole in terms of its constituent parts. Its encoding in DNA could also provide great insights, but both DNA and the brain are still hugely complex structures that were not designed to be readable by humans. It's possible that the whole thing gets cracked before 2029, sure, but this is the kind of prediction that people have been making for years, and it's always 20 to 40 years in the future, just far enough away to make it plausible. Usually the promising research at the time doesn't quite deliver and other projects and approaches take the spotlight.

Re:Billions

[IgnoramusMaximus]

Not only that, the GP (as many AI enthusiasts do) forgets that the synaptical connections are electro-chemical, not just purely electrical, and thus a whole new dimension of chemical communication enters the fray, complete with different functions of different neuro-transmitters at different synapses of the same cell, which can alter the functions of the said cell both short-term and long-term.

The more fair comparison to a neuron is not that of a transistor or even a logic gate but to a whole complete embedded microprocessor with up to 50 thousands I/O channels!

Each cell!

Now multiply times 100 billion...

Re:Billions

[Spy Hunter]

I think it's you who's forgetting something; namely the power of exponential growth. Try doing the math yourself. In 20 years, if Moore's law holds, a supercomputer will have enough transistors to dedicate tens of thousands to simulating each and every one of the brain's 100 billion neurons. The chemical interaction of two neurons may be complex but it's not magic; it can be simulated by a finite number of transistors. Personally I find it hard to believe that number is larger than the low thousands, especially considering how much faster transistors are than neurons. Even supposing it requires millions, that only pushes the timeline out a few more years. Unless you're disputing Moore's law, then arguing that the brain is too complex to even think about simulating it in a 20 year timespan is just silly. (Again, it's a valid argument to dispute Moore's law; it's got to end sometime after all.)

Re:Billions

[IgnoramusMaximus]

That is wishful thinking in the extreme. If any such technology comes along, it will not be transistor based. The very notion of a system composed of 100 billion times hundreds of thousands of transistors, no matter how streamlined and efficent is akin to someone in 1930s waxing lyrical about how in 1990 we will for sure have this mega-computing-machine composed of billions of vacuum tubes housed in 20 100-story buildings downtown New York and having its own hydro-elecric dam to power it. Moore's law will guarantee it!

Transistors are simply woefully inadequate for the job, very much the same way vacuum-tubes are.

Also "Moore's law" is nonsense. The growth of certain technologies, particularly at the stage of infancy, can be sometimes indeed exponential. But soon some hard limits are reached and the progress slows down radically. Just as is the case with modern x86 CISC CPUs, which having reached the boundaries of reasonable cost vs. performance are now essentially floundering in desperate directions involving simply multiplying the number of CPUs in hopes of some meaningful gains of performance.

To put it differently: we do not have at present any technology even remotely approaching the capabilities required. Will such a thing be invented and utilized in 20 years? Who knows. Note that in addition to the hardware capabilities needed, we also have, for all practical purposes, no clue as to actual functioning of most of neural systems, never you mind any insights into the nature of consciousness and how it relates to these functions.

But then again we were promised that all these problems will be solved 40 years back. Then just a little longer ... just a little ... HAL9000 was supposed to be zooming about to the moons of Jupiter in 2001, no?

This and commercial nuclear fusion, also "just around the corner" for the last 60 years or so.

Re:Billions

[Spy Hunter]

You know, you could have just said "Yes, I am disputing that Moore's law will hold for the next 20 years". That would have been a much more concise way to make your point, yet just as effective. And if you had bothered to fully read my two previous posts you would have found that I already explicitly acknowledged (twice) that Moore's law is anything but a certainty, and there was no need to construct ridiculous strawman arguments and vacuum tube analogies in an attempt to convince me of such.

If any such technology comes along, it will not be transistor based.

The very notion of a system composed of 100 billion times hundreds of thousands of transistors...

Transistors are simply woefully inadequate for the job

Listen to yourself; you're making the same predictions as Kurzweil, just in the opposite direction. There's no physical law preventing a supercomputer with quadrillions of transistors. There's no physical law preventing such a computer from simulating a biological system such as the brain. Your statements are merely opinionated predictions of our future nanotech fabrication capability, or lack thereof, in exactly the same vein as Kurzweil's, and with no more backing (somewhat less, I should say).

And while you rant about how people have been promising fusion for years, let me remind you that people have been predicting the end of Moore's law for years too, and so far they've been just as wrong.

Re:Billions

[IgnoramusMaximus]

Listen to yourself; you're making the same predictions as Kurzweil, just in the opposite direction. There's no physical law preventing a supercomputer with quadrillions of transistors.

Neither there is a "physcal law" preventing a computer made from quadrillions of vacuum-tubes ...

What prevents such a pipe-dream from occuring however are practical concerns.

There's no physical law preventing such a computer from simulating a biological system such as the brain.

We do not know that for certain. We do not know if a traditional Turing machine, which all our computers are in effect, is capable of such activity. Some models of neuronal activity for example are positing possible influences of various quantum phenomena at synapse level. Furthermore, our understanding of the nature of consciousness is pretty much at the level of the Ancient Egyptians' understanding of astrophysics. More neuroscientists dig into the neural cell operations, more complex things get. And that does not even address the intercell connectivity mechanisms, dendrite and axon formation, etc and so on.

Your statements are merely opinionated predictions of our future nanotech fabrication capability, or lack thereof, in exactly the same vein as Kurzweil's, and with no more backing (somewhat less, I should say).

Kurzweil is just another in a long line of prognosticians who finds his desire for media exposure and satisfaction of his self-importance being well served by stroking our collective egos and by appeals to hubris.

I am merely pointing out that the obstacles are truly paramount when viewed from the level of our current knowledge and available computing power.

And while you rant about how people have been promising fusion for years, let me remind you that people have been predicting the end of Moore's law for years too, and so far they've been just as wrong.

I am not sure what are you referring to. Moore's "law" fizzled out nearly half a decade ago when all the CISC CPU makers run into the physical limitations of gigahertz range frequencies in very high density integrated circuits. They have been all groping around since then attempting to parallelize whatever can be parallelized in hopes of achieving something akin to their prior advances in performance, results of which are highly questionable from the point of view of the silly "law".

Re:Billions

[Spy Hunter]

Neither there is a "physcal law" preventing a computer made from quadrillions of vacuum-tubes... What prevents such a pipe-dream from occuring however are practical concerns.

The practical concerns of a vacuum tube computer of such scale are obvious; on the other hand the practical concerns of a transistor one are not at all. The problems are almost entirely of nano-fabrication; unlike the vacuum tube case, the resources required are not at all prohibitive. There may be insurmountable practical concerns we do not yet know, but for you to proclaim they exist is no more defensible than Kurzweil's position that they do not.

We do not know if a traditional Turing machine [can simulate the brain]

Perhaps, but we *certainly* don't know that they can't. In fact we have zero evidence that brains do anything that can't be simulated by Turing machines, while on the other hand the pile of brain activities that Turing machines *have* done is ever-expanding (useful vision seems poised to fall next).

IMHO the likelihood that the brain performs quantum computing is small; however even so, Turing machines *can* simulate quantum phenomena to any desired precision. If you want to say that brains can't be simulated, you're going to have to do better than that. You need something like a "soul" that is completely inscrutable for it to be out of reach of Turing machines.

I admit consciousness is still a mystery, but our computers don't necessarily need to be conscious to exhibit human-level intelligence for useful tasks (and it may turn out that consciousness naturally "falls out" of any system complex enough to support it; I know it sounds lame but not more so than any other explanation of consciousness I've ever heard).

Moore's "law" fizzled out nearly half a decade ago

Um... surely you are not under the mistaken impression that Moore's law predicts clock rates? If so, you ought to visit Wikipedia. Moore's law is and always has been about the number of transistors on a chip. It has not failed to this day; the roadmaps look good for years, and the research looks promising in any number of new directions. Furthermore, you must not be following the latest in processor technology; innovation in that field has moved to GPUs, which unlike multicore CPUs are having great success in utilizing every transistor Moore's law can throw at them.

Re:Billions

[IgnoramusMaximus]

The practical concerns of a vacuum tube computer of such scale are obvious; on the other hand the practical concerns of a transistor one are not at all. The problems are almost entirely of nano-fabrication; unlike the vacuum tube case, the resources required are not at all prohibitive.

You seem not to grasp the dimentions of complexity and connectivity in human brain, thus your silly insistence on some magical "nanofabrication" (yet another pipe-dream nowhere near practical application) solving all the problems. Squeezing gazillions of transistors into some box is only one tiny aspect of the problem, dealing primarily with reducing the size of any AI to manageable levels. All the major unsolved problems remain untouched even if such a thing is accomplished.

There may be insurmountable practical concerns we do not yet know, but for you to proclaim they exist is no more defensible than Kurzweil's position that they do not.

This is hogwash. All I am saying is that given the current state of affairs combined with history of the field (something you are desperately trying to pretend does not exist) renders Kurzweil's "prediction" of full-fledged AI (and that means consciousness - no weaseling out of this one) astronmically unlikely to come true in the time-frame he specified.

As I repeatedly pointed out, he (and you) are yet another in a long, long line of uncritical, hubris driven propagandists of "just around the corner" great scientific "breakthroughs", which - quite mysteriously - fail to materialize decade after decade despite of billions upon billions of dollars sunk into the research.

Turing machines *can* simulate quantum phenomena to any desired precision.

This pretty much ends any discussion with you on the subject. You should explain how are you planning to achieve this feat to the physicists of the world and collect your inevietable Nobel prize in the process, given the current state of knowledge of quantum phenomena whereby the running joke is that there are more theories of these processes then there are physicists involved. We do not have a clue what is going on at those levels, only wild guesses, never you mind "simulating" any of it.

Fuck, we can't properly simulate much more higher-level processes involving whole atoms and molecules for lack of computing power and understanding, or did you not hear of this project which attempts to harness the power of millions of home computers to attempt to simulate, in a matter of weeks and months folding of a single protein!

You are out of your mind if you think we have (or will have in 20 years) anything like the computing power or knowledge required to simulate 100 billion neurons. Forget 100 billion, in fact we can't properly simulate one cell in its entirety and the prospects look rather grim that we will be able to do this in many decades.

If you want to say that brains can't be simulated, you're going to have to do better than that. You need something like a "soul" that is completely inscrutable for it to be out of reach of Turing machines.

We do not know the nature of consciousness, therefore we do not know if it is an emergent property of certain complex systems such as those that can be simulated by Turing machines, or if it is somehow tied to the quantum phenomena and requires something "extra" to be present such as a particular arrangement of molecules in some way sensitive to certain quantum effects the nature of which we do not yet understand. This does not necessarily preculde a conscious AI from being constructed but it would render the problem orders of magnitude more complex.

And then there is of course the possibility that consciousness is somehow dependant on some hereto unknown extra-dimentional entanglements which would throw the whole thing into complete

Re:Billions

[Spy Hunter]

Wow, you have quite some bile built up there. Moore's law has always been about the number of transistors (despite clueless journalists getting it wrong, don't blame Moore's law for their faults), it *is* still holding, and, crucially, it *is* still producing exponential gains in the performance of GPUs, which again crucially *are* translating into exponential gains on real-world non-graphics problems. GPUs are becoming more general-purpose all the time, while retaining the ability to translate more transistors into more performance. A whole industry has sprung up around using GPUs for general purpose computation and the fruits of that labor are already being put to practical use today. I've written such an application; have you even used a GPU before? Perhaps you would be interested to know that Intel is producing a GPU, due out in 2010 (engineering samples by the end of this year), whose major component is, guess what, general-purpose x86 processor cores with expanded vector units. By 2015 or so the distinction between GPUs and CPUs may be so blurry as to be irrelevant.

did you not hear of this project which attempts to harness the power of millions of home computers to attempt to simulate, in a matter of weeks and months folding of a single protein!

I was beginning to think *you* hadn't heard of it. Did you not notice that Folding@Home is in fact using Turing machines to simulate quantum phenomena? Randomness is trivially added to a Turing machine, producing a probabilistic Turing machine. With the simple addition of a hardware random number generator computers can simulate all the probabilistic quantum phenomena you want. And, incidentally, did you also not hear that Folding@Home has been able to realize large performance gains by running on modern GPUs? Those very things you insist are useful for nothing but graphics?

You should explain how are you planning to achieve this feat to the physicists of the world

I believe you are deliberately misunderstanding me, but let me clarify: [probabilistic, i.e. with a hardware random number generator] Turing machines can simulate any quantum phenomena [that we have a theory for]. It's possible that, at the bottom, quantum physics is unknowable and we will never be able to construct a coherent theory of it. It will be a sad day for science if we discover this. I tend to think this is not the case, and quantum phenomena have a coherent theory. Furthermore, I believe, though this is just opinion, that we already know all the quantum physics we need to understand how the brain works and it's more a matter of figuring out how the entire system works together. You have no real evidence to the contrary, so any speculation on your part is no more or less trustworthy than mine; certainly not enough so to calculate any "astronomically unlikely" probability of human-level AI in 20 years.

I think I understand what you are thinking now. You are extrapolating from simulating one atom, to one protein, to simulating an entire cell full of proteins, to simulating an entire brain's worth of cells. I think you are beyond foolish if you really believe that's what it would take to reproduce the exterior behavior of the brain. We don't simulate a car's movement by directly calculating the trajectories of oxygen atoms in the combustion chamber, and we won't simulate the brain's behavior by directly calculating the movement of individual amino acids in proteins. There will be massive simplifications at multiple scales during the transition from simulating proteins to simulating the brain. The behavior of proteins will be characterized in these exhaustive simulations, then modeled at a higher level with much less calculation. The behavior of a number of complete neurons will be characterized with detailed simulations, then distilled into the parts that are essential to calculation (possibly including some quantum effects; IMHO l

Re:Billions

[IgnoramusMaximus]

Moore's law has always been about the number of transistors (despite clueless journalists getting it wrong, don't blame Moore's law for their faults), it *is* still holding, and, crucially, it *is* still producing exponential gains in the performance of GPUs, which again crucially *are* translating into exponential gains on real-world non-graphics problems.

No it is not. I keep telling you something you do not want to hear: the rate of increase of the number of transistors on a single slice of silicone is not equal to the rate of increase of general computational performance! The two have long, long since diverged.

Furthermore, GPUs, by their nature, are specialized devices meant for very specific types of computations, which sacrifice many of the traditional abilities of a general-purpose CPU in order to gain speed in those specialized areas. If that was not the case ... they would have ceased to be GPUs!!! That is the whole point of a specialized piece of hardware.

Additionally, because GPUs deal with a specific type of computation involving large number of vertexes and texture operations which can be parallelized, an increase in the number of parallel pipelines of a GPU renders increases in computational performance. Which of course does not translate into gains in most other areas of general purpose computation.

GPUs are becoming more general-purpose all the time, while retaining the ability to translate more transistors into more performance.

Which is of course complete nonsense since such a trend would make them into general-purpose CPUs and would thus cripple any possible gains resulting from the necessary trade-offs I mentioned above.

The whole point of a GPU is specialization. That is where the performance advantages over a general-purpose CPU come from. Remove specialization and the gains evaporate.

Perhaps you would be interested to know that Intel is producing a GPU, due out in 2010 (engineering samples by the end of this year), whose major component is, guess what, general-purpose x86 processor cores with expanded vector units. By 2015 or so the distinction between GPUs and CPUs may be so blurry as to be irrelevant.

Which if it were what you are implying it is would result in a perfect circle of stupidity resulting in abandonment of all gains from specialization. Fortunately its just marketing babble. What you are confused about here is Intel's plan to roll the specialized GPU functions into yet another parallelized set of pipelines alongside a general-purpose CPU, i.e. more "cores" in one chip, except some of them with specialized GPU circuitry.

I was beginning to think *you* hadn't heard of it. Did you not notice that Folding@Home is in fact using Turing machines to simulate quantum phenomena? Randomness is trivially added to a Turing machine, producing a probabilistic Turing machine. With the simple addition of a hardware random number generator computers can simulate all the probabilistic quantum phenomena you want.

Right, and that is why the project is in dire straights because its results are consistently failing to correlate with real-life proteins. I actually had a talk with one of the people involved in the science of protein folding calculations and he indicated that all the serious research has all but abandoned computer simulations and is now proceeding by means of rapid, computer-controlled chemical experimentation. The reason? Insufficient representation of quantum phenomena (dealing with reactions with surrounding solvents) in the models results in their utter unreliability.

Using a random number generator is not the same as simulating quantum phenomena. It is a statistics based kludge which is used to estimate outcomes of quantum events. Not simulate them. Subsequently only models which do not depend on accuracy of representatio

*sighs* What to say ...

[golodh]

@timeOday

Has it occurred to you that all of us already work, to some extent, at the direction of computers? Think of the tens of thousands of pilots and flight attendants... what city they sleep in, and who they work with, is dictated by a computer which makes computations which cannot fit inside the human mind. An airline could not long survive without automated scheduling. Next consider the stock market. Many trades are now automated, meaning, computers are deciding which companies have how much money.[...]

That's enough. Err ... frankly your reply has given me pause. Seriously. It betrays a wealth of misunderstanding about AI and computing in general, and I have been wondering if I my reply should be a sarcastic one or just an explanatory one. Given the nature and the depth of the misunerstanding displayed here, I have settled on an explanatory one.

What you call "Automated scheduling" is part of a branch of applied mathematics known as "Operations Research". Basically it's the art and science of formulating a practical, real-world problem (such as air-crew scheduling, devising FedEx routes, loading aircraft, routing goods flows through transport networks as efficiently as possible, finding optimal stock portfolios, finding optimal ways of running an oil refinery, etc. etc.) into a mathematical problem, (usually a so-called "optimisation problem; see http://en.wikipedia.org/wiki/Category:Optimization_algorithms) and then devising appropriate solution algorithms that can be executed by a computer (usually a digital one) to give exact or approximate optimal solutions to said problem. See also: http://en.wikipedia.org/wiki/Operations_research

Such problems can be quite large ... e.g. with thousands of variables and tens of thousands of constraints. Now I'm confident that you would be quite unable to solve a 2x2 LP problem (i.e. a Linear Programming Problem, one of the most basic Operations research problems) in your head, or a 3x3 problem using pen and paper. Any PC can run a program that solves such problems in microseconds. This however has nothing to do with the question of whether solving an LP problem is to be classified as AI or not. As a matter of fact, solving LP problems is not, and has never been, considered part of AI. The same holds for all the other OR problems I mentioned.

Now it turns out that many of the problems I mentioned don't have what are known as "efficient" solution algorithms. Meaning we don't know of any exact solution algorithm that has polynomial run-time on a digital computer; instead all known algorithms have *exponential* run time on a digital computer. In such cases one resorts to what are known as "heuristics" (see http://en.wikipedia.org/wiki/Heuristics#Computer_science ), being algorithms that aren't guaranteed to find an optimal solution, but which sometimes *can* be guaranteed to come within say p% of the optimum, or at least to come up with a fairly decent solution. Some of the heuristics used, e.g. what are known as "branch-and-bound" algorithms (see http://en.wikipedia.org/wiki/Branch_and_bound) are based on questions that were (also) encountered or raised in the study of AI.

The important thing to note is that in general this has nothing whatsoever to do with Artificial Intelligence per se. Artificial Intelligence (AI) research on the other hand deals with problems like: "How can we induce computers to exhibit behaviour mimicking the Human Mind, or the Human body" (see: http://en.wikipedia.org/wiki/Artificial_intelligence))

Note the lack of overlap between Operations Research (OR) and Artificial Intelligence (AI) problems. The m

Re:*sighs* What to say ...

[timeOday]

It's the old receding horizon problem. One working definition of AI is that it's whatever we don't know how to do yet. Chess was the quintessential AI problem for decades. When superhuman chess performance was achieved, adherents of this definition said "oh, that's not AI, it's just search and table lookups." Well, maybe that's all there is. Search, regression, memory. If AI is understood to be "the magic part that only humans have" or "whatever isn't part of a more disciplined field" then eventually it will amount to nothing at all.

Re:*sighs* What to say ...

[Seraphim_72]

That is the most cognizant reply I have read. Kudos for you Sir, I doff my hat.

Reasonable estimate

[Anonymous Coward]

AI intelligence level's been beyond that of the average American for a while now, I suppose another 21 years to catch up with the rest of the world is reasonable...

No thanks

[namgge]

Machines that behave/think/emote like humans are a pointless vanity. Anybody who needs access to human intelligence and/or emotions can hire real people right now.

It's bad enough sysadmin-ing a conventional family. I really don't want to have machines throwing tantrums at work as well.

Namgge

Re:No thanks

[neminem]

Come on - you post to slashdot, and you voluntarily miss a perfect opportunity to reference HGttG? You even had your choice of references - either "Please enjoy your trip through this door", or "You don't have to pretend to be interested in me you know, I know perfectly well I'm only a menial robot".

A stupid competition, and Kurzweil won?

[Futurepower(R)]

I wrote the parent comment. Since I posted it, I've been trying to understand how Ray Kurzweil could say something so foolish as "We'll have intelligent nanobots go into our brains through the capillaries and interact directly with our biological neurons."

Not only is he saying that there will be artificial intelligence in only 21 years, but he is saying that the computers on which the new AI runs will be so small they can travel like cells in our bloodstream, and do useful work based on an extremely advanced understanding of biochemistry and an ability to interact on a molecular level.

There is no evidence that anything like that is happening. It is wild imagining.

I'm guessing that Ray Kurzweil understood correctly that the National Academy of Engineering Grand Challenges for Engineering is a publicity gimmick, and that the committee is a social group. Maybe Mr. Kurzweil decided to try to outdo everyone else in getting publicity. So, he put together the popular prefix nano- and the hot words robots, medicine, and AI. And he was successful. He tricked the BBC into quoting a prediction he himself doesn't believe.

Apparently, Ray Kurzweil interpreted the event as a socially backward macho male competition, and, given that, he won.

The National Academy of Engineering web page, Reverse-engineer the brain, is also wildly nonsensical, but somewhat more restrained, saying: "... further advances are needed...", and "Because each nerve cell receives messages from tens of thousands of others, and circuits of nerve cells link up in complex networks, it is extremely difficult to completely trace the signaling pathways."

There is lying, and then there is creative, energetic pseudo-scientific lying. There is treating other people badly, and then there is using a knowledge of science to take advantage of the shortcomings and weaknesses of other people. I suppose Ray Kurzweil was only getting into the mood of the baloney artistry the National Academy of Engineering created for him. But using baloney artistry to get attention is not only infantile, it is FRAUD.

This is all my opinion. If you can find a more positive interpretation of it, I'm interested.

Ray Kurzweil gave another interview about his imaginings that was rather uninformative, but not so nutty: Interview with Ray Kurzweil about the engineering challenges of the 21st Century (MP3, 6 minutes).

You have contributed to the tone of the discussion

[Futurepower(R)]

"Unfortunately the FSM forgot to provide us with the reference manual, ..."

It appears to me that when you involved the Flying Spaghetti Monster in this discussion you helped create exactly the right tone of seriousness that Kurzweil's statement deserves.

Maybe the nutty one is a follower of The Noodly One.

The Rapture

[4D6963]

Am I the only one who thinks that the Singularity sounds awfully like the atheist computer nerd version of the Rapture?

We will have both the hardware and the software to achieve human level artificial intelligence with the broad suppleness of human intelligence including our emotional intelligence by 2029.

And you base this claim on? Absolutely nothing? Oh, your wet dreams? Well that's different then..

Please allow me to predict the dream recording device for 2024, the first webcomic written by an human-like AI by 2037 and the infinite zoom algorithm for June 2031!

And the Rapture for 2048. I mean the Singularity.

human-level intelligence is an ever-receeding goal

[John_Sauter]

I don't have a PhD, but I was associated with the Stanford AI project in the 1960s. We made some real advances in Chess and Checkers, and did some interesting work on machine hearing and vision, but playing ping-pong was well beyond what our arm could do, and driving in traffic was well beyond our cart.

Our funding wasn't bad—we were able to purchase a PDP-6, which wasn't cheap—and the researchers had good imaginations. The problem of human-level intelligence, however, when you look closely at it, is a very large problem. We can define nearby milestones, and make some progress towards them, but I think creating a human-level intelligence is much more than 20 years in the future, just as it was in 1967.

AND!

[EddyPearson]

AND BY 2029, YOU'LL BE ABLE TO DOWNLOAD RICE.

I've rtfm, there is NO science at work here, just some bloke making HUGE unsubstantiated claims. They cite no research, they don't even make any concrete claims apart from "at a human level". I've seen more technical and in depth discussions between piss heads on a parkbench.

Finally at the bottom, they namedrop a google founder to try and make this sound more believable.

Shame on the BBC covering nothing, and shame on Slashdot for posting filler.

I have ABSOLUTLY no respect for

[killmofasta]

Marvin Minsky, He was going to have a PhD in Vacuum tubes in 1980s

"1967, Marvin Minsky: "Within a generation ... the problem of creating 'artificial intelligence' will substantially be solved."

Ray Kurtzwiel... compelety diffrent story. He is just not thinking outside the box? He is thinking the box has melted into a blue river, that tastes like straberries! He is the real deal.

I look at his work, his ideas and his predictions very carefully. He also has his critics, and others that dont like this style. You just cannot admit, that he is not one of the bightest minds of this generation.

Wrong conclusion

[houghi]

AI will not be raised to human levels. HI will be lowerd to the artificial one.

20 years

[sentientbrendan]

Futurists (whatever that means) have been saying that "human level" AI is 20 years off for well over 20 years now.

Anyone who's actually passingly familiar with the state of the field, and understands what rate the research progresses at (SLOWLY) can tell you this is bullshit.

Will computers ever reach human intelligence? If you mean, will they be able to do *math* faster than humans, it has already happened. If you mean, will you be able to have a conversation with a machine about how a movie made it feel, then certainly *not* within 20 years. In fact, I don't expect to be alive to see such a thing and I'm in my 20's.

One thing that people fail to understand is that AI research largely does not seek to make some kind of Asimovian robot, that interacts with the world like a human being does. AI is just another branch of computer science with exactly the same end, to write software that solves problems that people can't. There is no pressing problem that requires us to make intelligent androids to solve it.

Another thing that really annoys me, is that people assume that just as soon as computers are "complicated enough" they magically become intelligent, as if intelligence were some emergent property of complexity itself. This is utter crap science fiction thinking. Human intelligence is not some general power we have that emerges from complexity. Instead, what we call human intelligence is actually a bunch of separate systems designed to handle separate tasks, such as vision, sound, language processing, memory, deductive logic, inductive logic, spatial reasoning, etc. People often talk about their minds as a single indivisible entity with the semi magical property of being able to experience the world and reason about it, but this isn't really true. The mind is a collection of separate and distinct functional units each with its own task.

Each of these intricate systems evolved over millions of years, its design is not simple, and we won't be able to figure it out in a mere 20 years. We certainly won't be able to integrate them in a coherent way in that amount of time.

Some futurists, especially the ones that are obsessed with the so called singularity, will point at graphs showing that computing power is growing exponentially, and take that as some kind of proof that their wildest science fiction fantasies are going to come true.

Even if computing power increases exponentially forever (it *won't*, but that's another issue), it still doesn't solve the underlying problems of intelligence. If we have a thousands of cores of terahertz processors, our AI would only be marginally better. The reason for this is that we don't have a lot of the underlying *algorithms* for solving various intelligent tasks. It's not that we are too slow to solve the problem, it's that we either don't know how to solve the problem in the first place, we don't know how to formally state the problem we are trying to solve, or that we know how to solve the problem, but don't know how to solve the problem or an approximation of it in polynomial time. The fastest processor you can imagine doesn't really help with NP problems (assuming that NP != P, yada yada). Exponential growth in computing doesn't make the traveling salesman problem tractable.

Ethical AI algorithm, recursively hybrid

[aim2future]

I've read both quite optimistic and pessimistic ideas above about our possibilities to create strong AI. My idea is that it's doable, we have the computing power, and definitely around 2030 (in my thesis from 2003 I suggest 2030).

The problem I see is not the computing power, neither things like wetware like neural simulations where we have come far. The main problem is the architechture, to glue all this power together. I work with unsupervised pattern recognition which is one of the weak AI methods and from my view a key to AI. My strong belief about strong AI is that we need to design it using an ethical, hybrid reasoning, recursive approach.

• ethical, implies it needs to be programmed with love (or axioms expressing one for us ethical paradigm, like Asimov's three laws of robotics).
• hybrid, the only reasonable way to assure an ethical system I see as combining rule based reasoning with pattern based reasoning
• recursive, to make possible an arbitrarily level of intelligence, using a simple repeated structure, the system needs to be built based on simple modules which can be combined for arbitrary abstraction

The algorithm below may need some improvements, it's only conceptual, but within 10 years I believe that this can be implemented and as such work at any abstraction level within a system.
BEGIN
    Axioms := Load('Fundamental_Concept'); (* The axioms *)
    Goals := Load('Goal_Concept'); (* Goals as rules and hypotheses *)
    Priors := Load('Prior_Beliefs'); (* A priori beliefs *)
    Questions := Load('Questions'); (* Questions to be answered *)
    REPEAT
        Data := Collect('Data'); (* Data Collection *)
        Patterns := Inference(Data,Priors); (* Find patterns *)
        IF (Answered(Patterns,Goals)) (* Deduce goals *)
        AND (Answered(Patterns,Questions)) (* Deduce questions *)
        AND NOT Contradiction(Patterns,Questions,Goals,Axioms);(* Resolution! *)
        THEN BEGIN
            Proofs := ConstructProofs(Patterns,Questions,Goals,Axioms);
            Apply(OccamsRazor,Proofs); (* In case multiple solutions, simplest! *)
            RealWorldReport(Proofs); (* Report/use results *)
        END
    UNTIL forever;
END

As we want these AI to serve us, without really being dependent of us, if we, or they choose to escape this universe, I suggested this as the modified ethical laws:

1. Respect (love) your creator and competing life forms!
2. Strive to understand your creator!
3. Do what you can to fulfil your creator's desires!

That is, these creatures would have no choice but to love us, thus they wouldn't have free will. To create an AI that would learn to love and respect others, I consider a much too hard (and risky) problem that may take thousands of years to solve.

This may not create human like intelligence, even though it is insipired by introspection of my own thinking, but would we really want to create a creature mimicking our problems, taking into consideration that a large part of the human population have different problems with themselves, as power-hungriness, paranoia, anxiety, depression etc...? I think we create AI because we need assistance and to simply relieve us from tasks we consider too hard or too boring.

In our own case we are using a subset of this type of reasoning to implement a (patent applied) business method for AI-assisted customer driven innovation, but then we still speak about weak AI of course.

Precision

[SpaghettiPattern]

I like the precision with which an event that should have happened years ago -but didn't- is forecast.

Why not saying 'round 1930?

It's a bit like grandma asking for the time the umpt time and you respond with a millisecond of tolerance.

I say this everytime this comes up...

[Evil Kerek]

A computer can't generate a random number (yes I know they have faked it with arcing electricity and stuff) A computer is, as I hope most of you know, on/off values. Period. This will NEVER make an intelligence. Period. The only way we are ever going to make an AI (and GOD that word has been SO abused - these people are making expert systems NOT AI's) is to have a radical change in computer technology - and while we have some intrigueuing ideas, they have yet to be little more than a faster/smaller way to make bits.

It's a fantasy. All they are currently doing is making an expert system so fast it can fake intelligence..up to a point. The faster the computer, the more tricks you can build into it. But to suggest this is 'intelligence' - please.

I will add that a lot of this depends on what your definition is of 'intelligence' - and good luck getting a single answer to that.

And to those that suggest we have any clue how the brain works - all you have to do is try to deal with a semi-serious illness. Then you find out the dirty little secret - They really don't know how most of the human body works, much less something as comlex as the mind.

Just my .02
EK

And my P.S. Why, on a website as sophisticted and geeky as slashdot, is there no online spell checker? I suspect a lot of people here simply enjoy berating people for misspellings.

Re:I say this everytime this comes up...

[Anonymous Coward]

So you're suggesting that we have randomness in our brains? Although then you suggest we have no idea what goes on in our brains. We may be largely predetermined, right?

Re:I say this everytime this comes up...

[Evil Kerek]

I'm suggesting that a computer will never decide to suddenly stop working because it's bored, for instance. Or if I ask you to come up with a number between 1 and 100, the things that fashion that decision can't be accurately mimiced by an expert system because we really don't KNOW what affects it. No, of course it's not truly random either - but the factors that go into play on it are. Your feelings, the temperature in the room, are you posting to slashdot, blah blah.

My real point is that intelligence requires something besides a bunch of on/offs.

EK

Re:Where's my flying car?

[kklein]

Best post ever. Lately, I've just been kind of basking in the joys provided me by cheap oil: my computers, my video games, bananas in Japan. I don't expect any of these things to last. I wish I could muster the optimism some people seem to have about our ability to change things, but I don't think it's possible. I think we're at the apex before another Dark Age. It's gonna be a scary ride down, but for now, I'm all for listening to the fiddles while nature, human and otherwise, lays the kindling beneath our new Rome...

Logical fallacies

[Dan East]

I say this because thus far, we've discovered nothing whatsoever that is non-reproducible about the brain's structure and function

You got to the fallacy of your argument quite quickly - in the second sentence. We don't fully know how the brain works, thus there are likely entire mechanisms we haven't discovered yet. Are those unknowns reproducible? Who knows - simply to know of something does not mean you can create it at will. What we know about the brain so far has no bearing on what we find in the future and whether that will be reproducible.

The other fallacy is that we are held up by our current computing power. That's been a running excuse now for decades, and it's hogwash. I'm sure you know what it means to be Turing Complete - essentially any Turing Machine can emulate any other Turing Machine given enough execution time and storage. There are creatures with extremely simple brains. Put two and two together, and why can't we simulate a worm's brain?

This is from almost a decade ago - 1999:
"C. elegans has one of the least complex nervous systems of any life form on the planet," says Lockery, a University of Oregon biologist who has studied the worm for twenty years. "Its brain has only 302 neurons, or brain cells; that's compared to about a hundred billion neurons in a human's brain. It is the only animal for which we have a complete map of the brain. It is likely to become the first animal for which we can gain a fairly complete understanding of how the brain controls behavior."

So we have a full map of this animal's brain of 302 neurons for a decade. Well, where are the Flash and Java worm simulations floating all over the net, showing a little virtual worm moving around based on the stimuli of our mouse?

Re:Logical fallacies

[smallfries]

I hadn't come across that work before. They've made interesting progress in the past ten years. Although there aren't any simulations floating across the net they have reimplemented the network in silicon to drive their bio-buggy. Doesn't that count?

what bullshit

[sentientbrendan]

>An algorithm or algorithms is certainly required, and I never meant to imply otherwise.
>Human understanding of said algorithm(s), however, is explicitly not required.

What you're saying is that we don't have to understand how algorithms related to various intelligent faculties work, but we can just make it magically happen if we copy the structure of the human brain? And this will be *easy*?

You expect us to be able to implement an algorithm that we have no understanding of in a meaningful way? Or do you expect us to create an atom by atom simulation of the human brain? Both of these things are impossible fantasies.

>It is one thing to understand the mechanism required for operation --
>it is quite another to understand the state it is in. I think you are
>confusing the latter with the former; the former is relatively trivial,

Really? You think that the mechanisms of intelligence are simple? Maybe you'd like to enlighten the scientific community by telling them what they are.

The mechanisms are various structures and algorithms that are *phenomenally* complicated. For instance, the way we process language is incredibly complicated. Hint, it's not enough to just know the grammer, vocabulary, etc. You also have to understand semantics, context, before you can even *parse* a sentence (which is the opposite of programming languages where lexical and syntactic analysis can happen independent of and prior to semantic analysis). We are nowhere *near* being able to do this even in principle. Just "copying the solution" from the brain isn't an answer, as it is not possible to do that without understanding what the brain is doing in the first place. Studying humans helps, but it doesn't solve the problem.

>and the latter is not required any more than a complete understanding of the state of
>everything involved at NASA is required in order to create, launch and recover the space shuttle

If you need to make an analogy to something unrelated that you also don't understand how to do, it's a good hint that you don't know what you're talking about.

Re:what bullshit

[Khazunga]

>An algorithm or algorithms is certainly required, and I never meant to imply otherwise. >Human understanding of said algorithm(s), however, is explicitly not required.

What you're saying is that we don't have to understand how algorithms related to various intelligent faculties work, but we can just make it magically happen if we copy the structure of the human brain? And this will be *easy*?

He's saying that understanding the algorithm and understanding the algorithm state are different things, and that you are mixing the two. I happen to agree with him.

We do need to understand the basic algorithm for an artificial brain. We may not be able to understand the actually working brain.

Imagine this hypothetical scenario:

1. We fully understand how a single neuron works and communicates with nearby units.
2. We possess "classical" computing power a couple orders of magnitude greater than what is required for simulation of an entire brain.

In this case, we could easily mimick nature, and simulate evolution. We'd get a working brain, whose structure we did not design and which could be as incomprehensible to us as wet brains are.

I'm not stating that we need that much computing power before designing an artificial brain. We'll get there much sooner, because the problem of interfacing our brain with our current computing hardware is nearly solvable -- i.e. crude interfaces are already being put into use, namely in the field of artificial vision and hearing.

Re:what bullshit

[jdoeii]

> Or do you expect us to create an atom by atom simulation of the human brain?
> Both of these things are impossible fantasies.

Fantasy? Yes. Impossible? Don't be so sure:
http://en.wikipedia.org/wiki/Blue_Brain

> we can just make it magically happen if we copy the structure of the human
> brain? And this will be *easy*?

Yes, very easy and with pleasure. Millions of intelligent beings are produced every day without full understanding of the process. I am sure you even heard of it :-)

> >confusing the latter with the former; the former is relatively trivial,
>
> Really?

Yes, realy. Read any recent articles from Maass, Izhikevich, Sejnowski, Markram, or any other researchers on recurrent neural networks.

Here is another fact for you to think over. Human beings are encoded as DNA. The total size of the human DNA is about 400MB. That encodes everything, not just the brain.

> You think that the mechanisms of intelligence are simple?

Yes.

> Maybe you'd like to enlighten the scientific community by telling them what they are.

Scientific community is enlightened, don't worry about it. All you need to do is review say the last 5 years of research and you will be enlightened too. It shoud not take you more than a few months.

Re:what bullshit

[fferreres]

>>Fantasy? Yes. Impossible? Don't be so sure:

It does still look like fantasy.

Imitation AI without understanding what is truly happening is like trying to build Windows by taking a photograph of how the screenshots look, instead of understanding why thing move, how, for what reasons and purpose, and based on what principles. If you don't know what the program is doing, you end up coping widgets and maybe, if you are lucky, with a nice GUI drawing program. This is an analogy on.

And trying to copy the brain structure is like trying to "discover" Windows (or Linux) by replicating the hardware. You design an atom by atom medium (really simplifing things 10X and forgeting about 50K other things like many assembler ops, the bus, etc. that real hardware is doing) to what you think is the important part of the original Windows. An then you expect it to develop intself into ... Windows! Or maybe it's start at DOS, and develop into Windows...after all, Windows is like DOS but better, just Humand are like rats but better. And....Windows appears to reproduce itself with ease. Even Windows can copy Windows, it's only 400MB, about the size of DNA...just like humans!

And when nothing happens...you begin to recognize you need all the hardware for it to make sense. At you are then 1000 years into the future util you are done. But the next IBM, 50 years for now, will think that maybe, just maybe, if we populate the right conditions to the not-so-complete-platform (ie: really incomplete system, because if either understand the core stuff and what is REALLY happening , or you don't at all), say reading 1kb blocks of Memory of Windows every 10 minutes...then you have Windows. Will that ever fly?

No try switching Windows XP with Humans. Think about it like 10,000,000 complex hardware (operating on nature at a quantum level) and 1,000,000,000,000 more complex Software. And though in 100,000^5 of redudancies and things about nature, phisics, etc. that we don't know (because everything in Windows we do know, because we built it), and then add 1000x things we can't be aware of...a we are begining to grasp why unless you understand what's happening, you will end up with nothing much than food (as in proteins, carbohydrates and stuff that does nothing interesting).

Now, fast forward 200 years, and imagine you have the hardware, and the power to emulate the physical later, and that you can fully grow a person in-digit (as in-vitro but virtual). What you have there? A digital human! But you need to also replicate the world atom by atom for the human to be able to do anything interesting (abstracting and pretending the human doesn't need organs, and many interactions like receiving sunlight, food, etc.... means you are back at having to understand most every part and interaction to design for this...if that the case, go back to my previous paragraphs and re-read). So you end with a Digital Human that's a 100% copy of the physical layer. Suppose you could do that, and based on DNA you could grow a human digitally...SO DAMN WHAT? You gained nothing. You ported the human to an identical platform bit by bit. Yes, a Digital Human. You ended up doing an emulation layer that just wastes time and energy. You are better of doing a good service to that lady you saw yesterday that was in need.

All in all...I did not study physics, I did not read any of what you have posted, and I am not a computer scientist. But for crist sake...unless you understand what is happening (and you need to understand the code for that, and fairly well), or unless you understand exactly what is happening with the living thing (ie: dump of windows memory bit by bit, each hert)...and only then, you will be able recreate intelligence. Also, being able to emulate a human (but imitating nature) adds nothing to the equation. And no, it will definitely not be more able to make itself more intelligent that we are by self evolution...than what we are know. And my neighbor or coworkers, and even myself, could not evolve our DNA in any conscious direction, just as your digital version.

All in all, it's just more complicated than it looks, and that's my take.

Re:what bullshit

[sentientbrendan]

>Imagine this hypothetical scenario:
>1. We fully understand how a single neuron works and communicates with nearby units.
1. No we don't. We have some vague idea of how they work that is roughly modeled by a neural network in AI. However, neural networks aren't a magic solution to intelligence because there are different flavors of neural networks for different tasks both in the brain and in computer science, and all neural networks have to be trained.

Some neural networks in the brain are probably hardwired, and their configuration is determined genetically. We have no means of determining their configuration and mapping it to an electronic neural network.

2. even if we did have a perfect understanding of neurons it wouldn't be sufficient. You are talking about how the *hardware works*. A brain neuron is about as low level as a transistor. Nothing about a transistor tells you how to processor language or do anything higher level.

Additionally, there is no way to determine what algorithms our brain uses. There's no way to get a big graph of what neurons are connected to other neurons with what weights. There is no way to tell how the brain *updates* those weights (not all neurons are the same, not all follow the same rules).

>2. We possess "classical" computing power a couple orders of magnitude greater than
>what is required for simulation of an entire brain

The human brain has 100 billion neurons that form a graph. The average number of edges in the graph is phenomenal. Each edge is weighted and usually directed in most neural network models.

All of that is the *easy* part.

What is the configuration of the graph? There's no way to get that from looking at a human brain. This isn't fucking star trek, we can't look at anything atom by atom, cell by cell.

Again, *we don't know how the cells work*, so we don't know how to update the graph.

What I'm trying to impress on you, is that the idea of simulating a human brain cell by cell is impossible, and that's not how AI works, nor will it ever be.

Judgment Day

[Anonymous Coward]

3 billion human lives ended on August 29th, 2029. The survivors of the nuclear fire called the war Judgment Day. They lived only to face a new nightmare: the war against the machines.

WHY NOT SOONER ?

[posys]

Ray Rocks

Re:There is no such thing as Artificial Intelligen

[SpinyNorman]

There is a lot of talk about computers surpassing, or not surpassing, humans at various tasks - does it not bother anyone that computers don't actually posses any intelligence? By any definition of intelligence you'd like? Every problem that a computer can "solve" is in reality solved by a human using that computer as a tool.

Did humans design our own DNA and brain architecture?

Why are you willing to attribute to a human the intelligence exhibited by our evolutionarily designed wetware brains, when you wouldn't afford the same to a computer exhibited by its human designed silicon brain?

That's a very broad definition...

[mario_grgic]

of algorithm. It has been proven that there are things human brains do that are not algorithmic (or even computational) in nature.

Re:That's a very broad definition...

[fyngyrz]

Authoritative cite, please. Otherwise, no, it hasn't.

AI by 2029?? Nah

[mario_grgic]

I'd be happy if Vista had stable drivers by then :).

AI is not a hardware problem

[mario_grgic]

It does not matter how long does it take to compute a thought (if thought is computational at all). Is is a software problem (if it is tractable). So, if your company has something that can do this but it takes a 10 years to compute, rather than a millisecond, that is indeed huge. However, I seriously doubt that you do.

Re:AI is not a hardware problem

[SilentBob0727]

In the case of chess, the algorithm required to solve it is well-known and complete.

All it would take now to run the algorithm is time and memory. A LOT of time and memory.

It is well-known that neural networks are Turing Machine-equivalent. So, a neural network big and complex enough to solve "easy-for-humans-hard-for-computers" problems is easily implemented. It's just a serious pain in the ass to run.

Re:AI is not a hardware problem

[SilentBob0727]

I should say, however, that if P=NP someone actually proves it, "human-level" AI may be more within our reach than is currently thought.

Not that I'm holding my breath.

Re:AI is not a hardware problem

[mario_grgic]

Chess however does not equal to AI. Nor does it even require AI, as brute force inspection of each move is enough to play it well. As you point out, time becomes a limiting factor in that case and some smarts (on the chess playing algorithm writer's part) are needed.

token billionaire

[Anonymous Coward]

I always think it's funny to see Larry Page in the same room with folks like Craig Venter. He's always presented as a "big thinker" just like the rest of the crowd he likes to hob-knob with. Hmm, ya think it could it be your money and not your brain that folks would like to pick, Larry?

There's something very distasteful about the way so-called "intellectuals" constantly lick themselves around people with money.

Not about funding...

[mario_grgic]

The real answer is that it has become a religion for a log of "singularity" followers.

A very expensive baby?

[mattpalmer1086]

I hear these predictions that we will achieve "human level" intelligence due to exponential growth in hardware capability and simulation of neural structures. I don't have a fundamental problem with that. My only observation is that it takes humans at least a decade or two to get up to a reasonable level of useful intelligence, through interacting with the world and studying.

There seems to be an assumption that just having the equivalent processing power automatically creates a useful intelligence, with no time given over for learning and developing. So, are we talking about the raw capability to emulate a human level intelligence, or to produce a fully functioning intelligence you can interact with?

Otherwise, we're going to have a very expensive baby, that may or may not, turn out to develop into an interesting intelligence...

pfft, I'm pretty sure

[Anonymous Coward]

we'll achieve it by 2028. No more, no less.

linear vs exponential expectations

[constantnormal]

One of the flaws in the human neural engine is the inability to imagine/estimate/project a probable future (via virtualization or any other model you want to employ) that is based on a compounding of change, wherein the rate of change increases over time.

In the Real World, a chaotic behavior is achieved via a rate of change that changes, increasing exponentially for a time, with an eventual turn downward that leads to a chaotic collapse. But people seem locked into the use of a model that projects a future based on a linear projection of current trends, which works in a broad range of situations, but only for a brief time.

Time and time again people "project" their image of the future, whether a localized, short-term future involving the interaction of a few elements over a brief period of time (where this simple technique works fairly well), or a distant, global future involving many many elements interacting (where it doesn't work at all), continually underestimating the likely future that will unfold.

Yet this is exactly the way that most events progress.

We see this pattern of weak expectations in all kinds of areas whenever disruptive technologies enter the picture, from financial analysis to science fiction. And the people most afflicted by their inability to see the chain reaction of events multiplying to produce radical change in a shorter period of time than anyone would have believed -- those are the experts. And the exponential changes always seem obvious in hindsight.

Hopefully, whatever higher form of intelligence emerges in the decades to come, it will not fall victim to this fundamental flaw that persists in human intelligence. If so, it will have to re-invent itself to get around the problem, which may be an extremely intractable one.

One would have expected natural selection to have produced at least a few minds capable of projecting an exponential future, as a more accurate estimation of the future in the medium term would offer a profound advantage. But we have not yet seen such an anticipatory genius emerge yet, at least not from any history I am aware of.

AI by 1980 !

[mbone]

I am old enough to remember reading AI predictions from the early 1960's (see the subject for this post). I think in general predictions from 20+ years ago underestimate developments in computer hardware and greatly overestimate developments in AI, so I see no reason not to predict that the same will hold for 20 year predictions from our time.

I happen to think that we do have a super-intelligence that greatly exceeds any human intelligence - it's called science, and it is and has for decades been a
human-machine symbiosis. That will continue to become more effective as time passes, regardless of the state of machine consciousness.

Just in time for the 2038 issue

[Doug52392]

So once the AI are made, will they blow up when a bug like the 2038 UNIX time bug hit?

if you combined Eliza and a speech recognition ...

[constantnormal]

... engine, would anyone mistake it for "artificial intelligence"?

Hell, YES -- it would be all over the news, especially if it included one of those animatronic faces and projected some pseudo emotions.

But it you took a random selection of humanity and ran a Turing test between the two, how long would it take a typical human to tell the difference and correctly identify the "real intelligence"? And how often would they get it wrong?

The fact is that most of humanity fails the test if reasoning or logic forms any substantial part of it. Tell me, what part of "intelligence" is it that allows an entity to stand firm in their beliefs in almost mechanical defiance to reason and logic, when they cannot muster any response to an argument, and even admit that they are unable to do so.

Are humans, as a species, "intelligent"?

Examples:

* The universe is only 6000 years old. (never mind that they also believe in the things that make nuclear power possible)

* We are fighting the terrorists in Iraq to keep them from attacking us here. (never mind the fact that they acknowledge that the terrorists who attacked us had no connection to Iraq and that Iraq posed no threat to us)

You can add your own candidates to this list. They should be things that are widely believed, despite overwhelming evidence to the contrary.

What aircraft corner as fast as barn swallows?

[YetAnotherBob]

Any aircraft the size of a barn swallow.

Your question displays a lack of understanding. Not of biology, but of physics. Square cube law specifically. Aircraft don't corner as fast as small birds. the reason isn't any magic of biology, it's simple momentum.

The larger any object is, the more it weighs. Make it twice as big, it weighs eight times as much. packs eight times as much momentum. A large bird doesn't turn s fast as a small bird. Same is true of planes. Same is true of ships. A buss won't corner as fast as sports cars either.

A typical aircraft is 1000 times bigger than a swallow. It's a million times heavier. It packs a million times the momentum. It's not that the swallows design is better, or that there is some biological magic. It's just a question of size. It's true the other way too. A mosquito can turn a lot quicker than a barn swallow. Barn swallows catch mosquitoes because they can fly faster. Guess what, the aircraft you were so dismissive of can fly a lot faster than that barn swallow too. Visit a large airport. Swallows get killed by aircraft every day. They can't get out of the way in time. A barn swallow that was as large as a chicken would be ripped apart by the stresses if it were able to corner as fast as a real barn swallow. That's the real reason that chickens don't turn well in flight. (Yes, chickens can fly for short distances.) Momentum.

Your problem appears to be that you just don't understand scale. It is a wonderful thing when you do. You see reasons all around us, for all kinds of things.

So, yes, we should study biology. But, we should also remember the physics. The tricks the mosquito uses just won't work for a passenger jet. Nor will the barn swallows turns be good for the passengers on that jumbo jet. Still, some things will be useful. We just don't know what. Who would have thought that studying a sharks skin would help racing yachts. Personally, I hope that we get a lot of surprises. That's where the fun in science is.

I don't expect AI research to give us human type intelligence in a machine. Ever. That doesn't mean we shouldn't try. We don't know what we will get, or what it will make possible. We can't know before the fact. Studying birds didn't give us aircraft that can corner in a second or two, it did give us jumbo jets that can take us half way around the world in an easy chair. That took a lot of other things too.

The Wright brothers succeeded where Lilenthal failed. Not because they understood birds better, but because in the meantime the internal combustion engine was developed. AI will be the same. Right now, we don't even know what we need in order to make this work. There will be surprises.

Re:What aircraft corner as fast as barn swallows?

[ShakaUVM]

More importantly, we can't make a airplane the size of a swallow that will fly all day without refueling.

Re:What aircraft corner as fast as barn swallows?

[olman]

There's the issue of speed too. If you made a plane that could corner like a swallow when it's doing 1000km/h, you'd have to scrape the pilot from inside the cockpit class assuming the rest of the plane was made of magic material that wouldn't tear apart by the G-forces. So there's momentum and then we have inertia. Unless you have magic inertia damper most sci-fi warplanes seem to have you're not going to be able to change direction just like that..

Then again, a plane doing 7-8 gees can evade a missile (depending on your timing and if your deity of choice is feeling beneficial) a missile that can pull 25 gees because the latter is so much faster and actually has larger turn radius. And a small relatively slow sports plane can turn on a comparative dime compared to either.

Surprisingly that 25g missile is much much smaller and lighter than the plane as well.

Speculative BS

[kuzb]

They should quit speculating, and be like Nike. Just Do It. Everyone says "we'll have X by Y" few really deliver.

My gut feeling...

[CTachyon]

Warning: rambling post ahead.

My gut feeling is that, from strictly a hardware perspective, we're already capable of building a human-level AI. The problem is that, from a software perspective, we've focused too much on approaches that will never work.

As far as I'm concerned, the #1 problem is the Big Damn Database approach, which is basically a cargo cult in disguise. Though expert systems are useful in their niches, "1. Expert system 2. ??? 3. AI!" is not a workable roadmap to the future. I'm certain that it's far easier to start with an ignorant AI and teach it a pile of facts than it is to start with a pile of facts and teach it to develop a personality.

The #2 problem is the Down To The Synapse approach. This, unlike BDD, could quite possibly create "A"I if given enough hardware. But I think that, while DTTS will lead to a better understanding of medicine, it won't advance the AI field. It won't lead to an improved understanding of how human cognition works — it certainly won't teach us anything we didn't already know from Phineas Gage and company.

Even if we go to all the trouble of developing a supercomputer capable of DTTS emulation of a human brain — so what? If we ask this emulated AI to compute 2+2, millions of simulated synapses will fire, trillions of transistors will flip states, phenomenal amounts of electricity will pour into the supercomputer, just for the AI to give the very same answer that a simple circuit consisting of a few dozen transistors could've answered in a tiny fraction of the time, using the amount of electricity stored on your fingertip when you rub your shoes on the carpet during winter. And that's not even a Strong AI question. That's not to say that working DTTS won't be profound in some sense, but we know we can build it better, yet we won't have the faintest idea of where to go next.

That brings me to my core idea — goals first, emotions close behind. Anyone who's pondered the "is/ought" problem in philosophy already knows the truth of this, even if they don't know they know the truth of it. The people building cockroach robots were on the right track all along; they're just thinking too small. MIT's Kismet, for instance, gives an idea of where AI needs to head.

That said, I think building a full-on robot like Kismet is premature. A robot requires an enormous number of systems to process sensory data, and those processing systems are largely peripheral to the core idea of AI. If we had an AI already, we could put the AI in the robot, try a few things, and ask the AI what works best. So, ideally, I think we need to look at a pure software approach to AI before we go off building robot bodies for them to inhabit.

And how to do that? I think Electric Funstuff's Sim-hilarities captures the essence of that. If we give AIs a virtual world to live in — say, an MMO — then that removes a lot of the need for divining meaning from sensory input, allowing a sharper focus on the "intelligence" aspect of AI. Start with that, grow from there, and I can definitely see human-level AI by 2029.

Are we talking average Human?

[EdIII]

Maybe it is my cynicism... Maybe it is that I think Idiocracy is a documentary...

But when a good portion of the "average" Humans fail a Turing test 4 or 5 times in a row till they get it right, then maybe we should be shooting for a higher standard of intelligence.

I can only judge from where the world seems to be at this moment, but I would have to put average Human intelligence right now just slightly lower then a Speak N Spell. I say that only since a Speak N Spell probably "Speaks" and "Spells" better then the average Human.

How conveient

[taustin]

Very convenient timing. In 20 years, we'll have human level AI available to drive our fusion powerwed (available in about 20 years) flying cars *available in about 20 years) on the way to the spaceport for our vacation at the commercial resort on the moon (available in about 20 years), to receive our immortatlity treatment (available in about 20 years).

Yeah, right.

Re:How conveient

[cecom]

Man, thank you !!! I haven't laughed that hard in a long time :-)

Re:There is no such thing as Artificial Intelligen

[Livius]

And this isn't even considering the other side of the comparison. Artificial intelligence has nothing to do with natural intelligence, and we don't have a clue how the human brain really achieves (or is it simulates?) intelligence.

Here's how it will happen ...

[frogzilla]

First some clever people will model the neurons in some simple animal. Some good examples might be worms or ants. They will tweak that and play with it and learn a lot about how such simple networks lead to (what appears to be) complex behaviour. Then when they have that working they'll move on to a more complex critter, and so on and so on (some recent fiction has suggested lobsters as good early choice). Eventually they'll get to something really interesting. They will have a lot of experience to draw on from simple critters. They'll be able to run many different simulations, searching some large parameter space for patterns, varying parameters etc. Eventually the simulation will be equivalent to something smart and we'll ask it to think about its own parameters. It will soon be fine tuning itself. After this point we're probably all doomed.

Yeah, it's possible

[Anonymous Coward]

It's certainly possible, although that's a hard thing to say considering we're barely at mouse-level AI at the moment. As a practical matter, though, all we'd need is a thorough, full computatinoal model of the brain (at least one brain-mapping project is in the works, AFAIK) and the proper computing power, which we may have in 20 years, and we could simulate a human brain, thereby creating human-level AI. It may seem far-flung, but just think about how far computing power has come in the last twenty years--in 1988 there was barely any internet in the sense we think of it now, and computers were still clunky things with barely any memory.

Of course, people can debate whether this constitutes actual artificial intelligence or just a simulation of natural intelligence, but that seems like a pretty silly distinction to me.

Big deal

[TheLink]

OK assuming someone does create an intelligence by simulating a human brain. I still won't call that a breakthrough, unless you _understand_ the _why_ and _how_ , much much better than we do now.

Otherwise, if I wanted a nonhuman intelligence without really understanding stuff, I could always go to the pet store and buy one. Doh...

It's not too far fetched that the biotech people could create an "artificial" creature with human level intelligence (and before the Computer Science people, or "classic" AI people). By the AI field's current standards, wouldn't that still qualify as an AI?

That's why I think the AI field is still a dismal field - a lot of these AI researchers are doing _crap_, and many actually seem to think they're making significant progress when doing it. Sorry if I sound harsh, but hey even I can come up with bullshit that doesn't significantly advance the field.

The main advantage I see with the "simulate" approach over the biotech "grow it" approach is you could probably copy a desired training result more easily. Maybe that's all that counts - you could scale it better.

How about enhancing existing humans and animals with current technology, instead of trying to create a full AI? What problem are you trying to solve? Or are you trying to just create a new problem, or create a solution looking for a problem?

Not if we smash SKYNET in 2029

[up2ng]

2029 - Terminator World


* Advanced model T-800 with upgraded cloned skin made to look like specific people are introduced
* They can not be detected by dogs
* Skynets defence grid is smashed
* Both T-800s sent back
* Prototype T-1000 also sent back to 1995
* The Resistance sends a reprogrammed T-800 back to protect young John

It is easy to make predictions like this

[EmbeddedJanitor]

and far harder to deliver. "Flying cars by 2000" was not made by the people designing fly cars.

At this stage, AI is perhaps at the bug level or lower. Sure, for select tasks, AI can do some interesting things like fly flight simulators etc., but for "whole being intelligence" (what's needed to make an "organism" that can survive on its own, AI cannot keep up with a fly.

One more crank of the wheel..

[mstone]

The standard cliche is that human-level AI has been "about twenty years away" for the past 40 years.

Looks like Kurtzweil has fearlessly taken the banner and advanced that to "60 years".

Well ... yes and no

[golodh]

I agree that one can see a receding horizon effect in the sense that problems that were once considered to hold the key to providing AI (something that mimics the capabilities of a human mind (or body)) were later found not to provide much of an advance in that direction after all.

I submit that there is a fairly sharp criterion for detecting "real" AI: the Turing test. For a robotic equivalent of a human body this might be the tie-your-own-shoelaces test. Of course lots of subjects have emerged that could be studied in isolation and which yielded useful and definite knowledge, but were originally lumped under "AI" because people thought might help them on the way to create a "real" AI.

To me this mostly illustrates the fallacy of thinking: "If we could only solve this problem, we would have AI". So far it has turned out to improve our insight in what the Human mind isn't, and in parts (neural networks) how parts of it (probably) work. Both impressive achievements, but nothing in the way of constructing an AI that can e.g. pass the Turing test or tie its shoelaces. But those specialist subjects tend to lack the hype that surrounds "AI".

In this vein I think it's either an act of total irresponsibility to sell, what basically amounts to a wild-ass guess, as a "prediction" of the type the article mentions, or a deliberate attempt to use the "AI" hype with the sole purpose of obtaining funding.

Re: They predict human level AI every 20 years

[johnrpenner]

the thing that has never been explained to me though is -- how does a quantitative change (an increase in processing power) bring about a qualitative change?? ... que music -- just where is it that we say -- that's where the magic happens!??

2cents
j

far too optimistic

[Anonymous Coward]

fyngyrz wrote:
>
> Thirty six years from now, that ability to "simulate a few cells" should
> grow just in the *normal* scheme of things into an ability to simulate a billion
> or so cells without any trouble.


Well, let's say you're right and researchers are able to simulate 10 billion neurons in 36 years. There are 100 billion neurons in the human brain, with 10 to 20 billion neurons in the human cerebral cortex alone (a figure which does not include glial cells, which may also play important roles in cognition, and of which there are 10 to 50 times as many as there are neurons).

But, let's further postulate that given another 10 years (now we're up to 46 years from now) researchers will be able to simulate the 10 billion neurons in the cerebral cortex. Even then there's still a much harder computational hurdle left to overcome that hasn't even been touched upon in our calculations here. Namely, the interconnections and communication between the neurons. There are 60 to 240 TRILLION synapses in the cerebral cortex alone. And then there's all of the uncounted communication between the synapses via various neurotransmitters. How much longer is it going to take researchers to be able to simulate all that interconnection and the communication going on between the simulated neurons?

(See this link for more interesting statistical facts about the brain.)

Pruning

[tepples]

The only thing that humans do that AI doesn't (well) is automatically follow a few paths, rather then look at the whole picture. As an example, it has been shown (sorry no reference right now) that some chess grandmasters look only at a couple of moves and then calculate all the possible combinations from there rather then examine every possible move. This drastically speeds up the calculation, however it does miss moves that could be considered the "best". Isn't that what alpha beater pruning is supposed to do?

Re:Pruning

[teh moges]

(late reply, if you are interested) Alpha beta pruning works, but depends on choosing good first moves.

Human will become pets

[hackingbear]

If robot intelligence reaches human level and it will surely surpass human easily as it can be evolved and expanded much easier. What will happen?

First, many jobs can be replaced by robots -- janitors, construction workers, engineers -- will be replaced one by one. (Well... more likely it will be in the reversed order -- engineers, construction workers and janitor -- because computer will have higher logical processing skills than dealing with random environment.)

Then, jobs like -- scientists, CEO, Presidents, Directors of the Boards (the later don't much anyway and can be readily replaced even now) -- will be gone.

The whole supply chains of all products will be robotized -- even the cheapest labors from 3rd/4th/5th world countries will not be competitive.

Human politicians will be replaced by robot politicians. robot rights will be equalized to and surpass that of human rights.

At the end, human will become pets of robots, like dogs are pets of human.

Is that a bad thing? Probably not, see, my dog does not need to work but seems so much happier than I'm.

European or African?

[Anonymous Coward]

That depends: European or African swallow?

Nonsense

[gweihir]

This has been 20-30 years in the futire for so long, that the only sane guess is that it will stay there. Looking at the actual unerstanding of what intelligence is, the human rage still does not have a clue. True, it can be described, and at least some of us have it, but how it works on the inside is a mystery.

My personal optinion after having observed KI research for something like 20 years not, is that working, human-comparable KI is either infeasible with computers, or that we are more like 100-1000 years removed from achieving it. And then it is possible that it requires immense effort and has all the drawbacks of human intelligence, i.e. free will or at least a passable imitation.

Fixed that for you

[Anonymous Coward]

human intelligence will reach the level of Artificial Intelligence in around 20 years ...

Don't believe the predictions

[jasampler]

If someone builds an intelligent machine -even one just as intelligent as a non-human animal-, then nothing will stop the replication, reusing and reprogramming of that machine everywhere. Why choosing 20 years to reach such state? This may not be achieved by hard cooperative research, but the result of a fortuitous or accidental discovery having the ability of recognizing every perceived "object" and integrating dinamically those not seen before. Industry, military and government will do the rest.

clarification about atheism

[Dr. Spork]

I feel like I have to object to this characterization of atheism a bit. While I agree that many and even most atheists happen to feel passionately about various political and social causes (tending to be humanists and very concerned with improving human well-being), I think this ideological opinion doesn't have many analogies with religion.

For one thing, it does not compete with religion, and many strongly religious people (in every major religious tradition) have the same humanistic convictions and take their religion to support their humanism (and vice versa). The same goes for a belief in the results and methods of science: This belief does not crowd out religious belief, and most educated religious people in the West believe in science just as much as atheists do. Ditto for environmentalism and all the other ism's you mention.

You're right that various humanistic movements are organized, but so are chess clubs, national elections and universities. Belonging to an organized religion prevents membership in another organized religion (unless you're Japanese, who seem to have no problem with accepting several religions simultaneously), but it certainly does not prevent membership in another, non-religous organized movement.

I just want it to be clear that humanistic endeavors like the fight against poverty, for environmental conservation, for global justice, etc. are nothing like religions. Religion is a different sort of thing.

Atheists simply don't have a religion. What makes them atheists is that in them, any belief that gods of any sort exist, is absent. This does not force them to put their "faith" in any other movement in particular. I mean, to some extent, every human being with normal, human compassion has some sort of humanistic ideals. But again, that's just a result of being a moral and empathetic person, and it happens to moral people whether or not they have any faith in various gods.

Re:clarification about atheism

[Teancum]

I think this has touched a major nerve here, simply due to the number of replies I'm getting here.

The one aspect of atheism as a religion that is significant is that it does take the place of other religious thought within the mind views of those who profess such such a belief or lack thereof.

I also reject out of hand the definition here that religion necessarily has to require a belief in a higher being. There is nothing specific about environmental conservation that is necessarily religious in nature, but there certainly are fanatical individuals devoted to environmental causes that have turned it into a religion after a fashion. I'm not accusing all atheists of being an extreme environmentalists, but I am suggesting that this is a form of atheism after a fashion because God or a higher being doesn't play a part in this "belief". And those on this extreme fringe can be so fanatically devoted that they don't mind dying as a martyr for their cause, and these extremists to have a sacred cannon and other similar trappings of religion.

BTW, where did you get the mistaken notion that being an adherent to a specific religious philosophy necessarily requires exclusivity? Some (many) religions, especially organized "churches" (which can be distinguished from the religion) do demand exclusivity as a major tenant and aspect of their religion, but there are many individuals who don't depend on the viewpoints of a particular leader for their belief. Especially anybody with half a brain that is being honest about their beliefs. And not all religions require exclusivity either... even with formal membership and initiation rituals.

I'll be open here, I profess a general belief in the divine nature of Jesus Christ, and have a firm conviction that his teaching... if anybody actually paid attention to them for more than a passing thought... could do more good in this world by ending warfare and opening your mind to compassionate service to others. At the same time, I have also studied the philosophies of Muhammad, Buddha, Confucius, Isaiah and others. And frankly, from my own viewpoint, I belief that God in one form or another has touched the lives of all of these religious leaders. But that is my own belief here, and I'm adamant that others should be allowed to follow their own path to enlightenment or even morality in general how they see fit. Even if that means an out right rejection of religious thought altogether.

What is amazing me on this series of postings is a viewpoint that there is "one true atheism". Without a central leader pushing for exclusive control, how could there be with atheism? I'm not trying to define what you... as an atheist, believe or don't believe, but I am suggesting that even those who claim to be hard-core atheists to have religion... perhaps with a broader definition than they are comfortable at using.

And getting back to the business of talking about Artificial Life/Intelligence, the role of a soul is certainly something interesting to discuss in the context of religious philosophies... including a philosophy of a non-existence of deity of any kind. In that context, atheism is a religion in spite of claims to the contrary.

You mean genetic programming?

[marcosdumay]

If so, we absolutely need faster computers to proceed.

Non-deterministic != random

[marcosdumay]

So, please, clarify. I guess you meant random, non-deterministic computers are a theoretical thing that can make decisions based on future happenings.

Re:There is no such thing as Artificial Intelligen

[Anonymous Coward]

"There seems to be an underlying assumption that merely increasing the complexity of a computational device will somehow automatically produce intelligence."

Not quite. Their method of processing also has to change from line-by-line code to a dynamic path-based neural network, but researchers and developers are already taking that into account.

I can see why it's difficult for a lot of people to accept that computers could ever exhibit the same kind of sentience that humans do, particularly if those people are unwilling to acknowledge that intelligence and even consciousness is merely the result of a complicated stimulus->processing->action system. The human brain is a machine, albeit one that didn't begin its life as a tool for some other being (as modern computers did). There simply is no reason why an artificial system cannot achieve our level of sentience (which is really just a system of learning, remembering, and comparing) once it reaches an adequate level of complexity.

This may seem harsh, but I'd be worried about the entire future of AI if I knew that the people working on it were anything other than hard determinists. When you accept that the brain is a physical processing system, and doesn't bend or bypass the laws of physics to allow for "free will" or unpredictability on a physical level, then you're finally ready to start working on creating AI. But not until then.

We act according to the paths that have been developed in our brains, and every "choice" is just a determinable outcome based on the input we receive. The AI will do exactly the same thing. People with spiritual beliefs will never accept that the computer and its deterministic actions are essentially just like us, but so be it.

I'm sorry

[strange dynamics]

If what you're saying is legitimate, you can rest easy knowing I'm a skeptical jackass. However, I'm amazed that you have been modded up / taken seriously. Your post has all the hallmarks of any other snake oil pitch: coming soon, radical new change, can't talk about it because it's proprietary, no real information.

Although it's possible, it is not often that breakthroughs are made by lone investigators, secretly toiling on 5 volume book sets. The usual way these things work is with a record of peer reviewed (or at least demonstrable) advances. There is a saying somewhere about extrordinary claims and extrordinary evidence. Why are people taking you any more seriously than somebody who claims to have invented cold fusion or a perpetual motion machine ?

Re:I'm sorry

[Walt Dismal]

I fully understand your skepticism. I've been working on this since 1983 though, and it's all full grounded in strong theory. And I have some things I haven't mentioned yet because you wouldn't believe I can do them, without seeing them. if you think the preceding was grandiose claims, you ought to see what I have in the magic basket to come. But I haven't published because I didn't want to give away the show, or have to go back and correct early work as being in error, hurting credibility. I don't want to just hand 24 years of work to Microsoft for free. Once I have enough essential elements patented, I'll open things, as a coherent and viable alternative to current directions in the field. When I bring this all to market, it will open a whole new area of software. But it is not Open Source. If someone wants to build AIs, let them spend 24 years of their time, or license what I've spent a huge chunk of life on.

There are people of questionable validity in the field. For instance, John LaMuth has a patent on implementing emotion in software. His patent is full of handwaving - in one part he points to an expert system and says in effect "oh, the logic is all in that black box." But never explains anything about what is inside. I on the other hand have a legitimate and working alternative to both LaMuth and the more widely accepted OCC model of emotion in computing systems. OCC too has holes. My model not only works more rigorously than OCC, it passes the Occam's Razor test - it is clean and understandable and computationally efficient. It makes it possible for an AI to understand human emotion, and have emotions like a human.

Ask me some questions about people and psychology, I may discuss some details of my approach if they are not key pieces whose explanation would hurt my business model. Again, my speaking up was to counter the people who do not realize how close we really are to serious breakthroughs. I want people to understand that we are very close to a change of paradigms in both computing, and society.

Re:There is no such thing as Artificial Intelligen

[Anonymous Coward]

Every problem that a computer can "solve" is in reality solved by a human using that computer as a tool. By that definition, every problem that a human can "solve" is in reality solved by their boss, using them as a tool.

Artificial Intelligence at Human Level by 2029?

[The_Rook]

only because we've learned humans aren't as smart as we thought.

To summarize

[Jane Q. Public]

what I have been saying in this thread: The problem is that it does NOT quack like a duck. That is the main point I have been trying to make. We haven't even gotten them to quack like cockroaches yet, much less ducks.

I agree that if something quacked like a duck, and looked like a duck, and otherwise behaved like a duck in all significant ways (and not a cheap mechanical imitation), then I would be inclined to call it a duck. But we are nowhere near anything like that, at all. Not even a glimmer. And we are really no closer now than we were 30 years ago. THAT's why I say that his timeline is unrealistic. If the next 30 years at all resemble the last, then we will be no closer then, either.

we need quantifiable metrics

[DrEasy]

One problem is that we can't even agree on what it means to reach "human level intelligence", it's such a moving target! Decades ago, one big goal of AI was to beat the chess world champion, and well, mission accomplished! But then people argued that the use of brute force does not AI make.

Now people are working on soccer-playing robots, and already you can witness how in the RoboCup simulated league the agents are playing extremely well, and I have no doubt that with further advancements in robotics we'll see robots capable of beating the human world champions by 2050, which I believe is the goal of the RoboCup competition. But then I'm sure critics will say that such robots are merely being programmed for that one particular task, that the algorithms in question do not "exhibit" any intelligence, bla bla bla.

Despite all the objections I've been reading in this thread, AI has been extremely successful! It's everywhere - in your email spam filter, in your search engine, in your book recommendation site, in your sudoku-solver, in your classroom-scheduling software, in your face-detecting, smile-triggered digital camera etc.

We need researchers to agree on a formal roadmap of milestones to strong AI, with well-thought benchmarks and easily measurable success criteria (like "beat the chess champion", "predict the weather n days ahead of time", but unlike "pass the turing test"). This way we will be able to focus our efforts, measure our progress and have the ammo to respond to the nay-sayers.

Church-Turing makes be not believe

[shyberfoptik]

we may have sufficiently powerful hardware that the software can be "dumb", that is, just simulating the entire physical brain.

Unless there is non-Turing computation in that software, you'd have to accept that the same program being carried out with a pencil and a piece of paper would also be "intelligent." You'd need an extremely large piece of paper, of course.

I Believe, in 2029, it will arrive!

[Samarian Hillbilly]

Hey guys get real,
Theory proceeds practice. We don't even have a mathematical or theoretical model of either intelligence or consciousness. Let's take a sub-set, learning, we have computer learning, but do we have computer meta-learning, or meta-meta-learning. Wouldn't a turing machine that was capable of meta-meta...learning violate Godels theorem?
After all, such a machine could automatically create a meta-formal system that overcame the generated contradiction in the current formal system. Since the language the machine uses is itself a formal system, it should be able continuously generate meta-formal systems and thus disprove Godels theorem. We don't even have a non-recursive definition of consciousness. Definitions like "the feeling that you exist", pre-suppose something to "feel". Minsky's theory that consciousness is an illusion generated by sufficiently complicated systems, just begs the question, "if it is an illusion, just who is it fooling?".
I know the true believers are going to flame me for this.

Re:I Believe, in 2029, it will arrive!

[NoOneInParticular]

This argument presumes that a computer intelligence is necessarily a theorem-prover that tries to tackle the complexities of problem solving by deduction towards a fixed set of axioms. Your argument on Godel's theorem merely shows that a computer intelligence is not going to be a deduction machine.

Theorem-proving is about proving truth, optimality. Intelligence is usually about getting something good enough, on time. Completely different beasts, even though much of AI still fails to see this.

Re:I Believe, in 2029, it will arrive!

[Samarian Hillbilly]

Precisely my point, theory must preceed practice, if they are predicting an AI, what will be it's basis? Fine, it's not a Turing Machine, but until someone can say what it is, even if we don't have any idea how to do the nuts and bolts, all predictions are sheer fantasy, or to be more accurate, faith.

artificial reasoning != human intelligence

[lpq]

While I'm sure computers will be able to "out reason" us in the future, they will be unlikely to excel in human thought or creativity.

Let me try to explain: what concepts or thoughts can computers conceive other than what has been programmed by a human into them?

Sure, you can claim computers can "learn". But again, what value can a computer assign to any experience without having been told its value by humans?

The big problem I see for computers to develop into self-actualized beings is that they are not alive. Thus -- they cannot be killed, they cannot be 'hurt' in the way we feel pain. They can be programmed to say "ow" and act "like" they are hurt, but because they are machines, they cannot learn by avoiding painful experiences. One might think a computer might want to avoid being "turned off", but "why?" It can just as easily be turned back on and be none the worse for it. It may be better (may have been upgraded, may have had solar cells recharge its batteries...etc). There can be no fear of "death".

As such, there can be no desire to exceed, or grow beyond "death". As far as implementation -- how can computers "grow" new neurons or attachments. Our brains make new cells throughout life. Old cells die -- in a computer this would generally be considered a fatal error.

There are so many problems attempting to create human thought out of machines -- because they can never "think" beyond what values the humans have programmed into them. Or rather -- lets say this, I cannot think of anyway a computer could learn to "fear" something, like avoiding falling because a scraped knee hurts. I don't see anyway they can ever have "preferences", other than what they've been told to prefer. Could they prefer a shade of red or a flavor of a banana? How can they acquire all (or any) of the experiences that make us humans?

Maybe I'm missing some real obvious solutions, but I've been cogitating about this for the past several weeks (on the coat-tails of reading G.E.B. and a minor chunk into "AI: a Modern approach". How does a computer come to know what is "good" or "evil" (assuming one would even want one's computer to think in such primitive terms).

It seems like there is a very large, and not decreasing gap between computer reasoning, and "thought" or "self-reflection". Can you even imagine a computer "meditating" to "quiet its mind", to help its creativity increase?

Have to stop here....my output units are fatigued. :-)

-l

Typical problem with religious people.

[jotaeleemeese]

They have to explain everything in religious terms.

Get over it: we have no "sacred books","prophets" or "congregations" in the same sense as religious people do. Under that most asinine interpretation any club devoted to any hobby is also a religion.

Wordplay....

[jotaeleemeese]

You are very adept at it, but all comes crumbling donw if one scratches the surface.

You mention some "sacred books", you are framing the debate from a use of language that is clearly chosen with care to elicit a certain emotional response.

The fact is that people that govern their life by logic respect very much other people's work that use logic and science as a way to understand the Universe.

These people don't consider anything sacred. If we would have tomorrow a different explanation that is plausible for a given phenomenon, thinking people would be prepared to abandon the original works of orthodoxy and refer to the new ones, not because they have been revealed in a mysterious way, but only because the arguments there contained can be followed up and demonstrated.

Try telling a Muslim that some part of the Quoran should be removed based in the most elemental logic. Or a Christian. or a Buddhist (women menstruating are not allowed in some Buddhist temples because they are "unclean". Shall I need say more?).

Keep playing your little wordplays. They may attract some in the peanut gallery but don't cut any mustard once checked properly.

Re:Wordplay....

[Teancum]

Perhaps I'm being a bit of a gadfly on this issue. The problem here is mainly differences of a point of view, and allowing some individuals to define terms in such a way that they can be somehow "superior" to others of a different point of view.

And again, somehow a few of those responding have been assuming that they speak "for all atheists" when I hardly doubt that at all. Oh I appreciate their sincerity, but because the definitions of the words being used by "believers" and "non-believers" (to use some other "charged" words here) are used in significantly different manners and in reality have very different definitions.

Still, none of this is dealing with the main issue that started this whole thread.... what is the worth of a soul in relationship to the development of artificial intelligence? What is there about some sort of consciousness that can be described in mathematical terms? Can a study of religious views about this topic allow some greater insight above and beyond pure computer science related research?

Instead, brazen defense of atheism has prevailed here without even a reference to Artificial Intelligence... which was the whole point of the article in the first place, or so I thought.

Hunger?

[jotaeleemeese]

What does that have to do with intelligence?

Honestly, every single organism feels hunger and feeds itself. You need zero intelligence for that.

Re:Hunger?

[MrHanky]

Motivation. To have human-like intelligence, you need not only to predict human motivation, you need to be motivated like a human, both to discern relevant from irrelevant and to give the impression of actually understanding and not merely calculating the other. Hunger is just one basic example. Much human experience is also derived from food, and it's experience we have in common. We know, without trying and without having built a fucking database, that some foods will go well together and some won't. And so we also know, linguistically, how to make jokes by combining food words that are ridiculous together (and thus English cuisine was invented).

This kind of intelligence doesn't make you super smart, but that's not the point. The point is to make the smartness human-like. You won't get there by looking at abstract logic.

"In around 20 years"

[frenchgates]

Can Slashdot please make a new category for things that are "around 20 years" away, like AI, Fusion Power, going to Mars, Hydrogen-powered cars, etc.

Human intelligence is not monolithic

[Arguendo]

Speaking of "human intelligence" like it's a single concept is error. There are dozens if not hundreds of different human subsystems in the brain, and many of them are independent of one another. Language, vision, and auditory processes can all go haywire differently and independently of one another.

So what exactly do they mean by "human intelligence"? A system that can finally translate language decently? A computer that could actually drive a car in LA traffic? A system that can play chess without a gigantic brute force tree search? These are radically different problems.

So if they're talking about a system that can solve all of these problems and then become the synergistic whole, feeling, emoting, nuanced human brain, then this is crap. Predictions about AI have almost always been utterly, utterly wrong, even for simple processes.

I am reminded of the famous Emerson Pugh quote: "If the human mind was simple enough to understand, we'd be too simple to understand it."

Close...

[Jane Q. Public]

That's not exactly the same thing. But close. Whoever it was had a nice animated graphical demonstration of their Quicksort-beater.

Assumption

[elguillelmo]

I'd like to point out the underlying assumption of most of the comments in this discussion. Namely, that the works of the brain are algorithmical, and that what it does is *just* computing a calculus. This idea, coming all the way from Descartes and Leibnitz, is of philosophical nature. It is as mysterious and scientifically unprovable as the opposite.

Who is being deluded here?

[Anonymous Coward]

I do have strong AI right now running on a Von Neuman class processor

Technologically improbable. Suppose I give the benefit of the doubt for a second, okay?

But I'm real, sane, and not a charlatan, and have explained my technology to my patent attorney. I expect to be hiring staff within two years.

Okay, that's now socially ludicrous. Sorry, I just don't buy it.

As a history lesson, remember that 386s were initially illegal to export outside of the US, for reasons of "national security". You're claiming that you've solved the Holy Grail of AI research that's eluded the experts for 50 years, *and* that you're still free to publish your findings within the private sector, without the military envoking eminent domain, national security, or any of the plethora of public and/or secret powers it possesses over such world-changing technology?

Remember, the US military has stated that even affecting the US economy in an overwhelmingly negative way is grounds for US military action. Inventing artifical minds is beyond the pale.

You're deluding someone. I just can't tell whether it's your readers, your stockholders, yourself, or all three at once...

Re:Who is being deluded here?

[Walt Dismal]

You're entitled to your doubts. However, please refrain from declaring I'm deluded, inasmuch as you have no genuine basis for it beyond personal assumption. "socially ludicrous"? I live and work in Silicon Valley, I deal with the semiconductor industry daily, we have tons of startups and it is common to deal with lawyers here to guard our developments. Come on, now. I'll hire staff when I have some solid control over my intellectual property.

"Inventing artificial minds is beyond the pale". May I point out that the military very much wants good AI. They're certainly a segment in my business plan. And there are very flaky AI proposals that get funded by DARPA and yet are on far weaker grounds than what I have. I think I'm standing strongly.

But thank you for the viewpoints, they're appreciated exercise.

Re:Who is being deluded here?

[Anonymous Coward]

Come on, now. I'll hire staff when I have some solid control over my intellectual property.

If it's as good as you say it is, it won't be your property for long.

May I point out that the military very much wants good AI. They're certainly a segment in my business plan.

Shrug. They'll buy from you so long as it amuses them to do so. If you try to withhold something that they feel they need, however, they'll take it, and silence you: by any means necessary. That's just the way military forces operate.

You have the advantage of living the US, which gives you the option to co-operate with the US military without having to choose between serving your country and staying alive. But the fact that you can talk about your discovery on Slashdot openly suggests that the military doesn't know what you've accomplished yet, or it isn't as good as you implied that it is.