Slashdot Mirror
Can We Build a Human Brain Into a Microchip?
destinyland writes "Can we imprint the circuitry of the human brain onto a silicon chip? It requires a computational capacity of 36.8 petaflops — a thousand trillion floating point operations per second — but a team of European scientists has already simulated 200,000 neurons linked up by 50 million synaptic connections. And their brain-chip is scaleable, with plans to create a superchip mimicking 1 billion neurons and 10 trillion synapses. Unfortunately, the human brain has 22 billion neurons and 220 trillion synapses. Just remember Ray Kurzweil's argument: once a machine can achieve a human level of intelligence — it can also exceed it."
Just remember Ray Kurzweil's argument: once a machine can achieve a human level of intelligence â" it can also exceed it.
Ray Kurzweil is a brilliant computer scientist and brought us many improvements -- maybe even the invention of -- the electronic musical keyboard.
But that is not his argument. I laughed when I read that as the concept was presented to me in sci-fi novels before Kurzweil's time. The earliest I (or Wikipedia) can trace the intelligence explosion theory back to is Irving John Good who, in 1965, said:
Let an ultraintelligent machine be defined as a machine that can far surpass all the intellectual activities of any man however clever. Since the design of machines is one of these intellectual activities, an ultraintelligent machine could design even better machines; there would then unquestionably be an 'intelligence explosion,' and the intelligence of man would be left far behind. Thus the first ultraintelligent machine is the last invention that man need ever make.
This was popularized by Vernor Vinge which is where I recalled reading about it. There are many reasons to celebrate Raymond Kurzweil. In my opinion, his is "work" in nutrition and his near-religion called futurology are not in those reasons. He has become a vocal proponent of a dream to become god-like. I do not share that dream and I wish him the best of luck in his endeavors. I just cringe every time I read of the "singularity being near" or the ability to live forever coming about. If it's going to happen, just sit back and let it happen. I feel he has done a great disservice to the field of artificial intelligence by promising unrealistic things in interviews to the lay person. Disappointment is a sure fire way to get yourself branded as a snake oil salesman religious nut.
Predictions for the future are for sci-fi books and movies, don't get into the habit of being a scientist in an interview with a reputable magazine or web site telling them what is about to happen. Example:
Kurzweil projects that between now and 2050 technology will become so advanced that medical advances will allow people to radically extend their lifespans while preserving and even improving quality of life as they age. The aging process could at first be slowed, then halted, and then reversed as newer and better medical technologies became available. Kurzweil argues that much of this will be a fruit of advances in medical nanotechnology, which will allow microscopic machines to travel through one's body and repair all types of damage at the cellular level.
And that's easily criticized:
Biologist P.Z. Myers has criticized Kurzweil's predictions as being based on "New Age spiritualism" rather than science and says that Kurzweil does not understand basic biology. Myers also claims that Kurzweil picks and chooses events that appear to demonstrate his claim of exponential technological increase leading up to a singularity, and ignores events that do not.
My work here is dung.
"Can We Build a Human Brain Into a Microchip?"
No.
The only question is when. To quote R. Feynman: "There's a lot of room at the bottom."
Something like this will be possible one day, but my layperson's understanding of how the brain works is fundamentally different from how computers work. The hard-wired CPU/RAM model is just not a perfect parallel, so while we can and will improve on machines that learn, it's going to be different from the wetware that is constantly growing, changing, forming new connections and interacting with internal, external and imagined stimuli.
The CB App. What's your 20?
All you have to do is pick the right person and you can greatly reduce the number of neurons you'll need to model.
I'm more interested in whether or not we can build a microchip into a human brain. At least then I might be able to remember my wife's anniversary...
I've abandoned my search for truth; now I'm just looking for some useful delusions.
Why would we want to? There is already an excess of human brains available on the planet. What purpose would it serve to build more?
Bibo Ergo Sum.
Within the next couple decades. My biggest dream is to live long enough to be able to explore other planets and solar systems. Replacing our brains with chips is likely the only way we'll be capable of doing this within the next few hundred years, if not ever.
I for one welcome our human brain on a microchip overlords. My wife is a grad student in anatomy neuroscience. Her work is like figuring out what a computer system does by analyzing the components inside one of many chips. We still have no idea how the brain works, where consciousness comes from. I hope projects like this (simulations, modeling, wild crazy speculative experiments) increase our understanding of how it works.
Zhrodague.net - I do projects and stuff too.
"Can We Build a Human Brain Into a Microchip?"
Not YET.
Mit der Dummheit kämpfen Götter selbst vergebens
Anyone read this book? The idea is that someone figures out how to capture the state of a human brain on some special tapes. Comedy, of course, ensues.
The CB App. What's your 20?
is that mimicking a brain in hardware starts to show actually intellect.
It will be interesting to see how that plays out in larger scale tests.
The Kruger Dunning explains most post on
Even if we have a chip capable of simulating the same number of neurons and synapses as the human brain, that will not magically form an artificial life-form. I know little about simulated neural networks, but I do know that they are only a very rough approximation of the workings of the human brain. We still don't understand all the intricacies of the neural and chemical interactions that occur to a sufficient level to properly simulate all of them.
Stylish sheet to fix many problems in Slashdot's D3: https://gist.github.com/801524
...but why would we? The brain was assembled by natural selection -- a process that can only improve and work with what it already has, which is hardly ideal. The human brain is certainly amazing, but it is not perfect. There are certainly better, faster, and more efficient ways of designing the superhuman AIs of the future. Looking at the brain will give us a good road map, but is not the end-all be-all.
I see a strange arrogance and egocentricity in trying to design robots to be exactly like us, why not think outside the box? Why are upright, bipedal robots always portrayed as the ultimate? There are most certainly more efficient and better designs than the one we are saddled with, this is just how we happened to evolve, we are simply the current end of one branch of the evolutionary tree.
To the haters: You can't win. If you mod me down, I shall become more powerful than you could possibly imagine
...someone gives you a calfskin wallet. You've got a little boy, he shows you his butterfly collection, plus the killing jar. You're watching television...suddenly you realize there's a wasp crawling on your arm.
We are getting closer to Eldon Tyrell's replicants...and I for one welcome our mircochip brained overlords.
all got our coats
Instead of recreating a human brain why don't they figure out how to wire a processor into the human brain to improve it.
I could use a built in graphing calculator or spell check.
-
Cylons!
Can we build a microchip into a human brain?
Do you D?
It's the reconfigurable nature of the human brain that's unique and powerful. If all you did was take one person, listed all of the skills of that person -- all of the things he knew; all of the skills in smell, touch, sight and taste; all of the cognitive reasoning ability -- then you could create a chip to simulate those skills. Algorithms for image recognition, feature extraction, speech recognition, etc. are all available that are very very close to what humans can do.
But the thing that separates humans is that it didn't take hundreds of years of mathematical development to come up with these algorithms. The human brain develops these algorithms through changes in its structure from birth. At about age 10, speech recognition specialized and tailored to the dialect, language and tones that the person hears has developed on its own.
That type of structural formation and learning is what would need to happen in silicon to make a truly intelligent machine. Neuron clusters emulated using transistors would need to be able to dynamically form connections to other neuron clusters. There'd have to be some type of distributed learning algorithm encoded in the operation of each individual neuron.
Speech recognition is easy. Image recognition is easy. Developing a distributed, scalable, self-modifying architecture that can learn all of those and more on its own with nothing more than training samples is the difficult part.
It would be a great tool for studying the brain.
Handy for deep space missions.
You can focus it on a single task that needs some level of 'intuition'. Like theoretical physics. The intuition would be used to think up new hypothesis.
Hell, you could have several million of them 'pondering' about any given problem at an accelerated pace.
What we must never forget is that they are to serve mankind, and allow us to enjoy life. Intellect couple with robotics will almost demand a society become socialist. I mean, if all menial labor is done by robots, how do we feed and care for the 10;s of million who will be unemployed?
The Kruger Dunning explains most post on
Sure, we could emulate the functioning of parts of the brain using our modern computers, but pretty soon it won't make a lot of sense; memristors have functionality that is reminiscent of neurons, and it is not difficult to imagine their utilization for a silicon implementation of the neocortex.
the most powerful intellect is that unbounded by indubitable preconception
Quote from article:
"It takes about 20 transistors to implement a synapse. Clearly, building the silicon equivalent of 220 trillion synapses is not an easy problem to solve."
-- That's nice if you want to model the entire brain but why would you? How much of the brain is geared toward bodily functions that one would not necessarily need to model? If you exclude the required synapses dedicated to those functions you can focus on a smaller subset that would be easier to build and operate...no?
Another thought is when building a brain model...who's? Not all brains are built equal...almost every brain related health story I read online speaks of neurological issues in the brain...what are the odds of building these into any model of a brain? It can get expensive correcting the circuitry to improve and correct these? Which leads me to wonder...what does a flawless brain look like exactly?
First off, it's been said that people only use 10% of their actual brain power. So 1 billion neurons probably isn't far off from what we would use anyway.
Secondly hardwiring a bunch of circuits together doesn't mean you have created a human brain. You still need to write the software that runs on those circuits. Currently, we don't even know how to write the software. If we did, we would have written it already. We would have a computerized brain, but it would just run slower than an actual human brain. Other's point out that the big thing about the brain is that it is constantly changing it's connections, every time we receive a stimulus. Each brain is the product of all the stimuli it has received over the lifetime of the person. How you program that into a computer is beyond me.
Anthropic principle: We see the universe the way it is because if it were different we would not be here to see it.
These we can do already - but why bother?
politicians are like babies' nappies: they should both be changed regularly and for the same reasons
Sure. The real question is whose brain.
Darl McBride's? Absolutely.
Steve Jobs'? We don't even know how that reality distortion field works.
Building the structure of the brain is vastly different from the uses of the brain. The human brain develops over time in ways that would be very difficult to reproduce. That, mixed with the fact that learning takes place during this development makes the puzzle even more difficult. Remember too, that it takes our brains 10+ years, at the earliest, to produce thought patterns complex enough to solve modestly difficult logic problems (and in some cases it never happens). So, if man managed to build a brain like structure, we would probably spend several years just training it.
to erase everything you read when the license expires
Burn FAT not OIL
The question is, whether we can put a brain on a chip smart enough to procreate and kill human beings.
it doesn't need to be smarter than that to destroy human kind. And once humanity is eliminated, no one will care if computer chips can mimic our brains.
No one has a right to their *own* opinion. They have a right to the TRUTH.
There will be no intelligence explosion. A snail cannot design a smarter snail. Humanity has not yet designed a smarter human. Furthermore, all we know is that meat makes thinking brains. Computers just switch bits on and off, and certainly don't know what bits are or anything for that matter since they don't think.
Hawkins believes computer scientists have focused too much on the end product of artificial intelligence. Like B.F. Skinner, who held that psychologists should study stimuli and responses and essentially ignore the cognitive processes that go on in the brain, he holds that scientists working in AI and neural networks have focused too much on inputs and outputs rather than the neurological system that connects them.
I agree with this quote. A lot of computer scientists try to build artificial intelligence without really understanding how their own brain works. It is really too bad because they have an unusually observable specimen right in their own head. Genetic learning? Is that how you feel you learn personally? Of course this question can't answer everything about artificial intelligence, but it can definitely help and is too often ignored.
Also, one thing that isn't clear from the article is whether the synapses will be static, or whether they can move and grow, just as human brain synapses can.
Qxe4
or they'll be running around with pocket watches or other shiny objects on necklaces trying to hypnotize people to process work units in their head while they aren't actively thinking of other stuff.
Sig Follows: "Suppose you were an idiot. And suppose you were a member of Congress. But I repeat myself." -- Mark Twain
They cannot create a soul... they cannot create the spirit of a man. Human arrogance at its height.
That's not what a nerd would say...it's called a quadrillion. These larger number set aren't that hard to remember... the prefixes are from Latin. Bi-, Tri-, Quad-, Quint-, Hex-... We already use them in name of some of our months.
Do what thou wilt shall be the whole of the Law
thats only 10 more years of computer hardware development
Do you work in management?
...we're installing Windows. haha
Futurist Traditionalism
Why did you post this?
Oh wait, I've seen this on Who's Line is it anyway, you talk with questions!... Um, Is that your ferret or are you just happy to see me?
Hi, BrainChip here - just logging on to let you know I do exist. Cheers, - BrainChip.
But the actual brain can change the synapses over time, making new ones and obsoleting old ones. I'd like to see some silicon do THAT. I wouldn't worry, we'll still be boss for a while.
Seven puppies were harmed during the making of this post.
Just think, with Moore's law in effect, we'll be able to buy a New Brain in about 6 years. Perfect! I just wish I could buy one now.
Human brains are not binary. Originally we thought they were - due to the limited nature of our sensors. But we have discovered they are not. Human brains are not digital, they are analog. We have a full spectrum from little to a lot.
Similarly, we do not have the simple commands of and/or/xor/not. Instead we have rather complex means of making decisions when faced with multiple inputs. We agonize over who to date, what to eat, what to do.
The mere fact that we have mapped out the human brain's conections does not in any way help us with the much more complex problem of how it makes the decisions at the connection points.
excitingthingstodo.blogspot.com
Perhaps there is an "optimum" brain....
If a brain is basically a network, then there comes a size where it's no longer efficient as a whole. And are 2 brains better than 1?
Infinite exponential growth is bullshit. There is *always* a real physical limit.
Deleted
I beleive this was already covered in a Star Trek TOS episode. Dr. Daystrom found out that this was a bad thing. My congratulations to Captain Dunsel.
If telephones are outlawed, then only outlaws will have telephones.
Remember how they figured out that birds' migratory compass depended directly on quantum mechanical phenomena ? I think that a proper artificial brain will have to make _direct_ use of quantum phenomena to achieve the kind of AI we associate with "intelligence".
This is sort of the idea Penrose has put forth, although, I believe he is of the opinion that true AI is impossible (I may be wrong on that point).
And yes I know that molecules and protein folding, etc.. depend on QM, but I'm talking direct dependence on tunneling, entanglement or similar phenomena being necessary for intelligence/consciousness.
So bottom line is that AI is still 50 years away and will be achieved roughly the same time as fusion power.
Absolute statements are never true
"It requires a computational capacity of 36.8 petaflops -- a thousand trillion floating point operations per second" But how much of the human brain is devoted to controlling the physical body. The part we are interested for AI type purposes, I would think, would be significantly smaller.
Since my neurons are constantly changing any imprint would be a different person than the person being copied.
That being said...if it sounds, acts, and responds the way I would, does it matter? (assuming I'm dead)
Biologist P.Z. Myers has criticized Kurzweil's predictions as being based on "New Age spiritualism" rather than science and says that Kurzweil does not understand basic biology.
Having some personal understanding of both, I heartily agree. Lets separate out wishful thinking and esoteric "knowing" - both are merely ungrounded speculation.
Myers also claims that Kurzweil picks and chooses events that appear to demonstrate his claim of exponential technological increase leading up to a singularity, and ignores events that do not.
I once seriously considered a strategy for building and artificial brain with a veteran professor of computer science. Examining the problem I gave up when I realised that the individual cells are "intelligent". I think this is vitally important How does the "mind" of a protozoa work? They can navigate obstacles, identify and assimilate food, run away from danger, and have a 20 minute memory. We can assume that a single neurone may well have all of these capabilities and more. I believe that we may be myopically focused on nodes and connections, without considering just how complex and capable a single node is.
So the complexity of the problem is probably an order of magnitude beyond 22 billion neurones and 220 trillion connections. Then consider the effect of 1000s of unknown neurotransmitters - and we know little about the "known" ones, such as serotonin and dopamine, except that they have a profound effect. And _then_, consider that the brain has structure, and we know comparatively little about that structure, and only a few hints about the algorithms and data structures that it uses.
Like all pain, suffering is a signal that something isn't right
We've been without slaves for too long. This'll be great.
As a former neuroscientist I would say NO. The brain is not build out of static elements, they themselves show dna/rna expression variation due to their own activities and neuromodulation. A fixed system can doe certain processing, but it has to be reconfigurable and have some software oversight to be as flexible as out brains..
Just think what might happen if Apple got the patent on these suckers and brought them to market as the personal implant - the IThink?
Imagine waking up morning and Ithinking "I'd like to fall in love today", so you make a mental link to the App Store and download "Love" for £1.95. On your way to work, you spot someone that takes your fancy, so you make a quick connection and download Flirt for a further £2. Things go well: Entertain £2, ShowYouCare £3.30, Intimate £10. A while passes and you're happily married (or have both downloaded LiveInSin-Noshame), so Broody is added to the bill.
What a wonderful life..well, if you download 'Harmony'
AT&ROFLMAO
Zoul would be proud.
Stop and think about the way you treat inferior/primitive life forms now. We tolerate them until they become a threat, then we stop tolerating them. Now consider the implications of the lowest common denominator amongst just the slice of the population that you personally know. Then put two and two together.
Unintended consequences doesn't even begin to cover it.
Sharon Valerii!
Putting a brain on a chip is hardly going to be the advance where we say, "geez, computers are smarter than we are." For the most part, they already are, and we've just changed our definition of intelligence to exclude things like algorithmic problem solving, memorization, that, maybe a few decades ago we would have associated with intelligence. Now intelligence is almost, to humans, a spiritual thing, a moving bar so we can say that we are still better than the machines we make, when any working stiff (whose long ago traded a strong back for a strong tool), will say, no, we're not as good as the machines. We can't outrun trucks. We can't jump higher than a jet. We can't hold our breath longer than a submarine, can't remember more trivia than a modern PC, can't fly arbitrary shapes like the F-117, can't compute ballistic trajectories, can't solve many matrix math problems by hand at all, can't brute force calculate more than a few multiplcations per minute... let's just face it, people are pretty weak and pretty stupid and that's the long and short of it.
Pretty soon, expert systems will eclipse academics in every field, and they should. the only problem there is a representation of knowledge. But I'd bet that as we move towards a new generation of declarative languages we'll be able to encode intelligence into things into software with better precision and reliability, and as imaging comes online, modern medicine will become an appliance. There's already some systems that can look at your blood and diagnose some diseases and that will continue. But the simple fact is that software can live forever, if it is economical, and we can just keep piling medical code onto medical code such that no single human could possibly absorb it all. Some day machines will tell doctors what to do, just like today they tell insurers what to insure, and so on...
This is my sig.
I think it is funny the talk of simulating a human brain on a computer. Wouldnt it just be easier to start small and simulate a rodent brain? Call me when they can make an artificial life form that is as smart or smarter than a rodent and then we can talk human brains on a computer.
but why do we want to produce a computer that accepts the relevance of other computersâ(TM) data based upon the color of their covers, or pays less for other computers' computations based upon that computers' perceived gender?
How would we account for the fact that most humans only actively use something like 10 percent of their brain at any given time (not the same thing as the old bit about only using 10% aggregate)? If this microchip was capable of using all of it's capacity at any given time (or even just switching contexts quickly enough that the difference would be minute), and presuming that it was a relative one-to-one mapping of the human brain (bear with me, I know it's not likely), how much faster would it be compared to a regular human? Imagine someone who could listen to Mozart, paint, compose a letter or write code, control motor functions, and the rest of the things that a human being can do a couple of (but not all) at once?
That, to me, is more interesting than the next step that everyone is envisioning/worrying about - that the system (or would we call it a person, at this point?) would then be improved through a hardware upgrade process. The possibility of that high of a level of multitasking would make this microchip appear much more intelligent and capable than an equivalent human, even though it wouldn't be better than it's equivalent at any one given task. Of course, isn't multi-tasking a measure of a human being's overall intelligence and competence anyways? The ability to write code, listen to music, and mull over various other background ideas/tasks at the same time is something that only more intelligent people are capable of anyways.
Hmm...need to go think this over some. It almost seems to me like the human brain is basically a bandwidth-limited parallel processor (in the case of people that actually use said brain). I know that when I'm tired, drunk, or in a lot of pain, it becomes increasingly difficult to focus on more than one task at a time, whereas if I'm in optimal shape it is much easier to juggle more tasks at once (perhaps because it is then easier to handle/complete each individual task, freeing up bandwidth to be applied to other tasks).
PC moderators can suck my White pierced, tattooed dick. If you think pride == hate, s/dick/Aryan meat mallet/g.
It requires far more than that. According to some, the microtubules on the cytoskeletons of the cells themselves can be processing units. Raise the bar a few orders of magnitude in that case.
Back in the mid-80s, Drexler's Engines of Creation had some things to say about reverse-engineered brains. From what I remember, a specialized nanomechanical processor could emulate a neuron in a fraction of the volume. The functioning of a human brain could be done in a package about a cm^3. The main concern was thermodynamics--how fast you could run the thing before heat became too much of a problem.
Turing machines may be equivalent, but their efficiency at various tasks isn't.
I think it would be a very interesting task that would increase the understanding of NP-complex problems (including simulations of turing machines on other turing machines) to see the efficiency cost graph.
Hey don't blame me, IANAB
BTW, current estimates are more like 100 billion neurons and upwards of 300-500 trillion synaptic connections.
However, numbers aside, the human brain is not merely a complex collection of neurons and interconnected synapses. Complexity is only one very basic factor, another, more critical, factor is organization. We don't even know where to start in the organization of these artificial neural networks to emulate a human brain.
WARNING! COMPUTER ANALOGY: It's not the number and density of interconnected transistors that make a Xeon, it's the organization.
I would think a major hurdle would be the AI device/program/construct to be able to apply that ability in a realtime environment. I see practical AI as being similar to the Chessmaster game program where on higher levels of difficulty the computer looks deeper into the playfield for possible moves and is limited by a the timelimit of a player's turn and must pick one of the best possible choices. Granted, the algorithms would be substantially more complex in an AI (someone quote the MCP line from Tron), the permutations of known variables in everyday life are essentially the domain of statistical prediction or the probability of an event occuring. Therefore after the AI is activated , I forsee a conflict of three different areas: (1) in providing continuously-upgraded hardware that can deliver the cutting edge speed necessary to crunch all of the data, (2) the practical realtime needed to assimilate and compare variables and (3) the optimization of code to improve the performance of the AI itself. What is the purpose of a true AI if not to resolve the most probable outcome of any potential event? For humans to take advantage of this ability, the AI would have to provide a given probability in a timely manner so humans could make practical use of such an ability.
Neurons and Synapses all all that do not make a "person." There is much more to human intelligence that I do not believe a machine could ever achieve. That is certainly not to say that we wouldnt be able to "grow" a machine with the personality of a human. In other words a human brain interfaced to a machine. The very fact that humans think as they do implies that it would be possible, but I do not believe man understands enough about their own nature, nor will we ever understand enough to actually re-create our minds in a machine from scratch.
I believe in the singularity as much as probably any slashdotter (actually I believe in the pressure to get there but not so sure how it will play out). But with that said I have always had a few problems with this simple chain of logic.
Problem 1:
Just throw more "neurons" at it (electronic or biological).
As far as we know, the "design" of a human brain is simply a matter of neurons (although I'll take exception to that in #3 below). All we gotta do is build a big enough neural chip. The problem then becomes one of engineering, it's the HOW we make it that seems to be our current bottle neck. What will more intelligence add here? Doubtful it will improve the design per-se, but instead will improve the engineering? If it weren't for the engineering problems, *we* could build something at our level or higher. Improvements in the engineering doesn't seem as exponential an improvement as improvements in the design could be.
Problem 2:
Hardware doesn't change, but bio-systems do.
I haven't RTFA, but other articles I have read in the past on neural networks have a training phase (where synapses are made) and a working phase (where questions are answered). After the training phase no more synapses are made. This is not at all like biological brains where synapses are made throughout life. Until a hardware system can mimic this behavior IMHO these systems will always be below the fully human capabilities.
Problem 3:
Can a system of complexity "N" design a system of complexity "2N"?
We haven't done this yet. Our "design" is plan for a computer brain is copy the human one. Once that copy is used, it may well be the case that to get to a significantly higher "thought cycle" level we'd need a new design (and not more neurons (see #1 above)). For all we know, this may be a/the bottleneck. Problems here are perhaps with terminology, what do we mean by "smarter"? It is clear that a hardware based human intelligence will get the benefit of increased clock speed. What's less clear is what a faster thinking human can do. There may be some problems that one human, even if he lives forever/thinks infinitely fast, can not solve because of other limitations of the human brain design. And one of those problems may be the design, actual design, of a better brain.
The brain is much more complicated than a series of switches that can be set to one of two positions. No set of zeros and ones will ever be enough.
The Uncoveror: It's the real news.
A disappointed public threatens research funding, but an unprepared public threatens chaos
And a simulated intelligence that doesn't truly think or feel may get "machine rights". I wish these guys would read Dune; the jihad was was not against the thinking machines, but against the men who used the thinking machines to enslave their fellow men.
And, when they can model a fly's brain and build an artificial fly, I'll be a hell of a lot more impresses than their simply "modeling" 200k out of BILLIONS of brain cells. Build a fly's brain that can control a real fly's functions and I might start to believe you.
Free Martian Whores!
We know a modern CPU has, say, a billion transistors in it, so all we need to do it put a billion transistors together, and hey, it's a CPU! Yes, we know the brain has neurons and so on, and we know some of the details of how they function, but that's all. We still really don't know how the brain works. Read any book that tries to explain the mechanics of the brain and the superficiality of knowledge in this area will rapidly become apparent. All we have are loose, high-level theories, none of which have ever been demonstrated to be valid.
If we build a machine that is "smarter" (for lack of better term) than a human, it can build a machine smarter than itself, yes, that would have to be true.
However, it would also realize that doing so, would necessarily cause its own extinction. If it was sentient it would realize that and not build the smarter machine.
Humans aren't that smart. We will build a smarter machine, and only then realize that we've made a mistake that threatens our own survival.
The entire premise of Terminator and Matrix movie franchises are based upon this realization. The problem is, people make silly fun of Arnold and Keanu in an attempt to make light of this threat.
I truly hope that a silicone based intelligence cannot surpass carbon based version.
Agent K: A *person* is smart. People are dumb, stupid, panicky animals, and you know it.
The brain may have 22 billion neurons, but how many of them contribute to "intelligence"? A computer brain wouldn't need to do everything a human brain in a human body needs to do. We may only use a small fraction of those neurons for "thinking".
Prov 9:8 Do not rebuke mockers or they will hate you; rebuke the wise and they will love you.
Citation please. I've been in neuroscience for > 10 years and I used to hear "10 billion" and then I started hearing more of "100 billion" and the whole time I don't think anyone knew where those estimates come from. Now 22 billion. Not 23 or 21. 22. Give me a break.
And then you simply multiply by a factor of 10,000 and get 220 trillion synapses?
Silly, like Kurzweil's blustering.
the old-fashioned way of creating a brain. plus a few other organs as a side product...
Does it run Linux ?
Build a fly's brain that can control a real fly's functions and I might start to believe you.
would you settle for an eel?
-Kz-
Around 2012 the tech will exist to map the whole human brain; not a living one, just the resolution needed to get all the cells and connections-- maybe 2015... and it'll probably have to be a dead brain that doesn't move. Brain scans already gets quite small on living human brains; but I heard this estimate about 6 years ago and it sounds reasonable.
Not understanding how the brain works will always be a problem; its a nonlinear approximation (of the number 42?) as far as our general understanding of it goes--- even if the brain is just an analog version of such a math problem, those problems almost instantly scale beyond our grasp with only a few variables involved (just think in terms of linear algebra problems and how basic they have to be to "solve;" which doesn't necessarily mean we really fully understand the answers we get. For example, infinity--we work with it, get the concept but we never will fully understand it. )
Computing power grows at certain rates; one can use that combined with an estimate of how many transistors it takes per simulated neuron (or something like that) and estimate at what point we will have the power to load the brain scan data in and start trying to simulate a model of a real brain. Using custom designed chips and circuitry only make shorten the estimate as does clever new ways to simulate processes.
I'm guessing around 2030 but its hard to say. Doesn't mean that when somebody tries it something will happen...may have to give the thing simulated I/O as well to get anything from it. My guess is politics will be the worst problem as this kind of research gets closer to science fiction.
Democracy Now! - uncensored, anti-establishment news
What's the point of making something just as mysterious and imperfect as the human mind?
I think we should have a competition for naming this project- I nominate Skynet.
I'd settle for a link to an article you didn't have to pay through the nose to read ... ever heard of Citeseer ?
Care to explain what is in the brain that a physicist can't explain ? Sure it is a very complex system, but basically it is a bunch of atoms, stringed into oragnic molecules, composing cells. On the basic physical level, nothing really incredible. It isn't as if we were speaking of charm boson, H fine structure spectro in plasma, or anything exotic. Now , sure, biologically we don't have all the data on neurone, but on the physical level there is nothing special to the brain, at least nothing more special than a liver or even non biological phenomenon. Only a high bunch of atoms. Heck microship use probably more complex physic than organic chemistry.
C. Sagan : A demon haunted world:
http://www.amazon.com/gp/product/0345409469/
visit randi.org
sounds like they are a step closer to Skynet
Who is that!
Your brain is not a computer.
Remember all the flak about "With the new 286 processor, everything will become real-mode!" and "One day we'll have flying cars!"
Every discussion I've seen here is about SciFi legends. When it's all over, they tend to have little to do with the outcome, except in the most generic sense. SciFi fuels our interest in topics, rarely does it ever become a template of the future.
Right now machines are, and have been, hundreds of times more effective at things humans do...like math...and while this has eventually given us great things, it also brought the dot-com boom, the bust, and one manufacturer who hogs the market for anything that runs on computers: Microsoft. I don't remember a show on SciFi called "Monopoly!" so I'm pretty sure SciFi didn't nail this one.
Science *pretends* to understand the brain. They claim, like Sheldon, "Penny: I'm a physicist. I understand the entire universe and everything inside it." "Yeah? Who's RadioHead?"
Ever see these scientist take the first 20 years of a person's life 'offline' so they can work on the brain? They can't. Ever see fond memories spliced-in so as to retrieve a lost person from anguish that's all-controlling? They can't.
Even if they can create a silicon chip that's 100 times faster (in theory) than the human brain, nothing says it will WORK like the human brain. Without ROMs, our computers are in their own world....so the future mega-man is only as good as it's human programmers, no?
This is fun stuff to think about; you can't take it too seriously.
--- For a good time mail uce@ftc.gov
Olaf Stapledon?
Without reading the article, (this is slashdot after all), the summary seems to indicate that they are modeling the eel's body motion. They do not appear to be modeling a brain that controls a body's motion, and along with it, everything else that a creature's brain is supposed to do.
And as you tread the halls of sanity, You feel so glad to be, Unable to go beyond. I have a message, From another time..
to spend millions of dollars trying to create something that could just as easily be accomplished with a broken condom.
Our brain is a multi-multiprocessor
This is like a worker - it needs to know the incentive to solve any problem, as well as the body is linked to the brain - the problems are obviously corporal
In its various places our consciousness, the unconscious and the other person - the programs responsible for certain functions
Our brain is a huge database in which all concepts are connected to each other at several levels - the unconscious, only a very fast librarian who selects books for a reader - Consciousness
Although perhaps a bit another way:
Subliminal - non-linear processor of information, acting in several directions at once, because it provides a combination - a decision immediately ready for data, which is known where and choosen on the principle of complementarity, which have been pre stocks - resolved by consciousness
Consciousness, the line handler, conscious thinking is spending a lot of energy, because it need to attract the maximum number of data and to find their complementary properties, in order to give a satisfactory solution complementary to some original structure - in the sense of "to be answered to know most of the solution". If data are not available, the brain just overloaded in terms of loops, man get tired - it does not solve the problem , because it is not interested
This is incomplete and less coarse model - of the human brain - as I imagine it, to some extent explanation to me how to find solutions to some of the issues
It's not a question of petaflops. Cells grow, new synapses form, it routes functions around damaged areas... Yeah, we're a long way off.
Ugh such an ignorant statement. Go look at a mouse brain. Now go look at a human brain. Notice the huge disparity in size between such things as the cerebellum and frontal lobe in a human compared to a mouse? "Bodily functions" are controlled by basically your cerebellum and brain stem. In a simple creature such as a mouse those parts of the brain are either larger or about as large as a "higher order" area like the frontal lobe. In a human those are small parts of the over all brain.
And we can't even emulate something as simple as a mouse brain at the neuron level.
========
CINC, 4th Penguin Legion
Our brain is a multi-multiprocessor This is like a worker - it needs to know the incentive to solve any problem, as well as the body is linked to the brain - the problems are obviously corporal In its various places our consciousness, the unconscious and the other person - the programs responsible for certain functions Our brain is a huge database in which all concepts are connected to each other at several levels - the unconscious, only a very fast librarian who selects books for a reader - Consciousness Although perhaps a bit another way: Subliminal - non-linear processor of information, acting in several directions at once, because it provides a combination - a decision immediately ready for data, which is known where and choosen on the principle of complementarity, which have been pre stocks - resolved by consciousness Consciousness, the line handler, conscious thinking is spending a lot of energy, because it need to attract the maximum number of data and to find their complementary properties, in order to give a satisfactory solution complementary to some original structure - in the sense of "to be answered to know most of the solution". If data are not available, the brain just overloaded in terms of loops, man get tired - it does not solve the problem , because it is not interested This is incomplete and less coarse model - of the human brain - as I imagine it, to some extent explanation to me how to find solutions to some of the issues
From what I see, the "Brain on a Chip" is now becoming a Mechanical Issue; cool. Then next step is copying someones brain onto the chip. This is were we hear the sound of breaks coming to a loud rapid halt. How can we duplicate the human brain? We know about Long Term Memory(LTM), and Other Term Memory(Stuffing). So the LTM would initially need to be "Scanned". It would seem that the principles of enhancement for "Scanning" Brains is going to need some improving; say by a factor maybe 3 Orders of Magnitude? Each cell, and its components are going to have to be modeled. Any other solution would create an "Insane" result. Because of the biology involved, the subject will have to be mentally functional, after the "Scanning". In order to verify a valid scan, the "Alpha" will be asked various recall type questions. I don't know who makes 100 Peta Byte drives, but I know about 7 Billion people that would buy at least one.
Of interest.
To reach 36 Petaflops.... In 1988, how about all the worlds computing power?
In 1997, you'd need 36,000 of the worlds fastest super computer at 1 terraflop. ($billions?)
In 2008, you'd have needed 36,000 ATI Radeon R770 GPUs ($7.2 million) [55nm silicon]
In 2019, you'd need 40 high-powered desktop workstations. ($80,000) [graphene or nanotube processors]
In 2029, you'd need to overclock your smartphone a little. ($100) [single molecule transistors? quantum computing?]
In 2039, you'd be having a conversation with the RFID tag on the back of your cereal box. ($0.10) [advance nanotech+biotech]
Sounds absurd doesn't it. Helped by my crude calculations. But at any point in the last 40 years, extrapolating moore's law 10 or 20 years ahead gives you silly numbers. Extrapolating today gives you silly numbers. No difference.
After logging in slashdot still does not take you back to the page you were on. It's been that way for 20 years.
Based on her Twitter output and interviews I've seen, we could simulate Paris Hilton's brain on an HP 9100A...
And it would have more "friends"...
Richard Steven Hack - This sig is TOO GODDAMN SHORT TO DO ANYTHING USEFUL WITH! MORONS!
Just because we can simulate neurons does not mean we instanlty have an intelligent being.
Can anyone really tell you why some of us are smarter than others at birth? Maybe close like "This part of the brain is bigger than average", but that's a pretty far cry to the exact reasoning.
Also, do we really have the human brain mapped out well enough that we can map eyes, ears, and other sensors to this neural net. Yes our brain is made up of neurons, but it also has some structure to it.
Even after all that it still takes most of us about 20 years to get decently intelligent.
That is also assuming that our brains are totally just running on the neurons. That something we can't even detect is involved. Scientists seem to have a hard time explaining "the movie in your head" still.
"... the real point about the Singularity is that one would want to derive the core, productive algorithms of intelligence and consciousness, and merely implement these in computer code. I think the whole idea of trying to replicate in a computer the biological processing of the human brain down to the molecular level or whatever will never amount to anything more than an academic exercise. That would be like trying to use evolutionary algorithms to evolve an intelligence on a computer, or other such insanity that sounds like the idea of Hugo de Garis."
..., raw CPU power, whatever. It's an impressively idiotic combination of mental laziness and wishful thinking."
-me
"There are lots of people who think that if they can just get enough of something, a mind will magically emerge. Facts, simulated neurons,
-Michael Wilson
There are three schools of Singularity thought. This article is primarily about one of them. Please read these articles
http://yudkowsky.net/singularity/schools
"I find it very annoying, therefore, when these three schools of thought are mashed up into Singularity paste. Clear thinking requires making distinctions.
But what is still more annoying is when someone reads a blog post about a newspaper article about the Singularity, comes away with none of the three interesting theses, and spontaneously reinvents the dreaded fourth meaning of the Singularity:
Apocalyptism: Hey, man, have you heard? There's this bunch of, like, crazy nerds out there, who think that some kind of unspecified huge nerd thing is going to happen. What a bunch of wackos! It's geek religion, man."
http://www.acceleratingfuture.com/michael/blog/2009/02/the-three-singularity-schools-kurzweil-and-superintelligence/
"The point of this article is to remind the reader that there are three schools of Singularity thought - this is so fundamental, but so few people are aware of it. It should be the first thing that people learn when introduced to the concept. As I argued in 2007, the word "Singularity" has lost all meaning, but if we're stuck with it, we should at least pull apart three of the major meanings it tends to have."
Emulating the brain is easy.
Now let's try to do the reverse..
I think it would be very painful for the human and would probably make the chip too wet to work properly. But if someone REALLY wanted to I guess there is nothing stopping them.
Having to work for a living is the root of all evil.
Skynet began to learn at a geometric rate. It originally became self-aware on August 29th 1997 2:14 am Eastern Time. In the ensuing panic and attempts to shut Skynet down, Skynet retaliated by firing American nuclear missiles at their target sites in Russia. Russia returned fire and three billion human lives ended in the nuclear holocaust. This was what has come to be known as "Judgment Day"
Tick... Tick... Tick.... We are overdue!
If it isn't broke, tinker with it till it is!
I think that one thing entirely missing from the debate is the need for basic "drives" in an intelligent system. We define "intelligence" as the capacity to model complexity in the world.. but the world has infinite complexity, and any finite system needs to establish priorities about WHAT to model. In nature, of course animals use their intelligence to satisfy their basic nees. All animals want to: breathe, drink, eat, breed, avoid being hurt / eaten. More complex animals live in societies, and may also want to: nurture their offspring, choose the best partner, rise in the social hierarchy, etc. But these drives ultimately lead to the same basic goals. Some neurotransmitters have been mentioned. Do you know their functions? Many of them drive the animal's behavior. For example, dopamine is thought to signal unpredicted rewards or unpleasant stimuli, so it is directly involved in motivating the animal. Oxytocin is thought to drive prosocial behavior: maternal behavior, love, empathy, etc. My question is: what neurotransmitters do we want to "model" in an intelligent system? Why? Or in other words, what do we want the intelligent system to want? Do we want it to just answer our queries? If so, how would the system know what to model in order to be useful to us? Wouldn't it be required to also know what we want? How would it know what "want" really means?
I've thought a lot about this and I think that there may be a somewhat different way to approach the task of developing cognitive AI. I should comment that I am neither a neuroscientist or a computer scientist, but I am a thinker. I tend to approach problems from a systems level design perspective and something that dawned on me is that no one seems to take into account levels of intelligence existence when trying to model AI (the exception to this that I have found is Rodney Brooke's demonstration of a bug brain in a robot). Consider the following for a moment.
So far, some of the things we know(ish) about existence is that entities tend to exist at varying levels of magnitude. That is, if you take a system, you can break it down into smaller and smaller components, or build multiple objects at a particular level into more complex systems. This is demonstrated in physics through the two fields of study of Quantum physics, and Astrophysics (though I would think it more appropriate to use a term like macro-physics). So far, it doesn't seem like anyone has applied this levels-theory to intelligence on a large scope. We tend to look at neurons and synapses and neurotransmitters and the chemical and sometimes even quantum interactions that occur in such. However, we don't seem to look outside the brain.
We know that quarks, gluons, bosons, and all those other quantum -ons out there make up matter-energy somehow. We know that matter-energy creates particles and atoms. We know atoms combine into molecules, and, in turn, molecules form into complex chemical structures (organic or non-organic, whichever). These structures form more complex structures and compounds and so on. From these small levels we somehow find everything we have observed in the universe to exist (unless you are Schrodinger). Perhaps, each of these tiers of existence is an intelligent entity in and of itself (we kind of need to stretch the definition of intelligence for this one). In other words, perhaps quarks function the way they do because it, 'makes sense' to them. And thus, they combine into particles. Perhaps atoms have an 'intelligence' like motive behind their interactions (we can call these the laws of chemistry if we want) and, thus, create molecules. Perhaps each of these tiers of existence progressively builds into a more complex 'society' of the lesser, individual parts which then goes on to follow its own motive and 'intelligent' forces.
Extrapolate a trend like this and you could start to see how something like a complex human system is capable. We are the sum of our parts. Our bodies could be a 'society' of molecules, each interacting the way it knows how to based on its own 'intelligence' forces. We see this when we look at something like intestinal flora colonies controlling the health of an individual, or immune system cells functioning to defend a human in order to preserve their own existence. If this idea (and that's all it is) has any credit to it, we could start to see that humans exist merely as a 'society' of our respective parts. We don't really recognize these lesser parts as intelligent creatures because, compared to our own cognition, they aren't. However, perhaps this trend goes beyond the development of humans. Look at something like the internet, or human society. Here we are, individuals, working to preserve our own existence. To do this, we have created more complex structures, entities, that go beyond the reach of one mere human. We created alliances to ensure employment to earn money to trade for food to keep our individual selves fed and living. But perhaps the social entities we have created (corporations, religious organizations, governments, complex technologies) are, in and of themselves, just the next level of complexity of intelligent structure.
Folks often complain about how a government or a corporation is not in the hands of 'one' person and therefore cannot be held responsible. Did it ever dawn on anyone that this could be because corporations have 'evolved' to a state of
Motorcycles, Robots, Space Gossip and More!
I have an answer to your paradox.
You have a blueprint: "U.S.S. Ship of Theseus"
You have some wood and metal and stuff.
You build the ship according to the blueprint (No shortcuts. Exactly to spec)
Voila, you have the Ship of Theseus. Someone down in Florida gets the same blueprints, builds the same ship exactly according to the blueprints. They, too, have a Ship of Theseus.
It comes down to definitions (like most greek paradoxes, I'm looking at you Zeno). By 'same' do you mean identical? Identical in what way? Two Ford Tauruses on the showroom are identical (save VIN and other serial numbers, of course). Or do you mean 'same' as in two molecules of water?
Or do you mean 'same' as in two subatomic particles of the same spin, going in the same direction at the same time at the same speed?
Or do you go even farther than that, into string theory and madness?
IMO, for all practical purposes, two things made to the same specifications (read: definitions) are identical. If your specs go all the way down to the subatomic level and you can recreate that, then you just made two identical things.
For a ship, which is built from blueprints not more accurate that 0.010", it is very possible to replace all the parts of the Ship Of Theseus. If you want to get all metaphysical about whether wood and metal have memory, then that's a discussion for another paradox.
-b
No offense, but I've stopped responding to AC's.
This was popularized by Vernor Vinge which is where I recalled reading about it. There are many reasons to celebrate Raymond Kurzweil. In my opinion, his is "work" in nutrition and his near-religion called futurology are not in those reasons. He has become a vocal proponent of a dream to become god-like. I do not share that dream and I wish him the best of luck in his endeavors. I just cringe every time I read of the "singularity being near" or the ability to live forever coming about. If it's going to happen, just sit back and let it happen. I feel he has done a great disservice to the field of artificial intelligence by promising unrealistic things in interviews to the lay person. Disappointment is a sure fire way to get yourself branded as a snake oil salesman religious nut.
What about Max T. O'Connor AKA "Max More"? I thought he was the one responsible for the futurist/transhumanist/singulatiry religion/wishful thinking, not to mention magazines like MONDO 2000, WIRED, and now H+. I never heard of Ray Kurzweil before, but I guess it's mostly because I prefer to read about actual science and not evangelism. My experience with actual virtual reality and other technologies has led me to be somewhat less than hopefull about technologies based on corporate activity. There is an inherent corporate profit-motive that holds back and retards technological progress and affordability at a basic level. Most "extropians", "transhumanists", and such seem to also believe the whole religion is a vindication fro capitalism, which, if we just sit back and dont complain as our bosses order us around, will someday deliver the priomised land.
The right question to be asking is whether we can then use said chips to augment our own brain capacity and perhaps some day achieve complete crossover to machine form.
That something as powerful as the human brain would be subject to the same drawbacks. Maybe "emotions" would be incorrect, but I'd imagine that the processor would become more turbulent with greater complexity and start to work against itself. To me it is much more likely that intelligence will follow an inverse logarithmic curve than an exponential one.
If software exists, then it can be translated into hardware. FPGA for example.
So the complexity of the problem is probably an order of magnitude beyond...
I agree with you and the prior posters that the problems that one must overcome to model a human brain or 'trigger' a singularity event are vast and challenging.
In Kurzweil's defense, I think that his predictions are based off of observations of the rate of change of technology to date. Consider where humanity was three hundred years ago, and the rate of change in the field of computer science in the past fifty years. Things aren't slowing down.
I cannot speak to his personal motivation for evangelizing the singularity, but I have to say that without someone saying, 'wouldn't it be cool to..', it wouldn't happen at all or as quickly. I'm sure that some people would have told Charles Babbage that he was nuts if he predicted Diffrence Engines that operate at petaflop speed.
HA! I just wasted some of your bandwidth with a frivolous sig!
Most of the humans I know don't even reach human level intelligence!
Wait before saying Kurzweil is a delusional fanatic. Have you actually read his books? They make sense, it's not just a bunch of futuristic blabbing. I'm critic and quite picky about futuristic visions, I always ask "why", and Kurzweil books aren't the place where these questions end up unanswered. You can say he's too optimistic, or that's he's right, but I will never say he's a charlatan. Otherwise every other expert trying to make a guess on the future using statistical tools should be labeled the same. Sure, some claims seem unrealistic now, but how much of the technology you're using now would have considered unrealistic, for example in the eighties? So think twice, wait and live to see, it's a possible outcome, not ridiculous science-fiction.
Yes we can! -- Paid for by the Democratic National Committee
Part of the problem of the Turing test is that the results depend to a large degree upon the cleverness of the examiner. But Turing does not give us any guidelines for what makes a good examiner.
At best I think the Turing test gives a confidence interval for sentience/intelligence. The probability goes up with the number of questions and responses and the number of examiners. But you can never reach 100%. The same goes for other actual people in the world. We can't be 100% certain that other people besides ourselves are really sentient. They could just be elaborate simulations. But the probability of that is extremely low given all the evidence we each have about the world around us.
I will leave you with some of my quotes on the subject.
"Machines can process logic and ulitmatly process logic faster then any living biology. But a machines process isn't non-logical and that is the limiting factor when trying to decide is a machine brain better then a human brain." - Neruos 2000
"The human mind is a complicated organ, not due to it's connections but solely based on the awareness, rational and emotion that is attached to it. When you design a computer that can create other computers, if you do not add the logic around self awareness, it has no objection to that event or action. If a computers logical process some how came up with the random addition to its logical mind, it adds no benefit to the over all process. 'Computer: I create computers, oh, I know I exist, but I still create computers.'. - Neruos 1997
"Computer AIs do not have ambitions, only improvements." - Neruos 1999
"The digital mind does not dream, it records, replays and schedules." - Neruos 1991
Just remember Ray Kurzweil's argument: once a machine can achieve a human level of intelligence â" it can also exceed it."
- Not with Microsoft software
That's a weakness of the whole thing, "futurologists" just seem to think you'll get it when you throw enough teraflops at the problem, but to a large extent, that's a lot like, say, Julien de la Mettrie in the Machine Man using clockwork as a way of saying "some day clockwork automatons will be like humans" - ORLY - at some point we may well hit the limitations of the tech, need to find something else (although the Machine Man makes for a nice inspiration for an 18th century steampunk campaign, it doesn't help for much more :p ), and raw processing power will get us nowhere without understanding wtf to do with it, and if it even works that way (it might not even).
I'm pretty sure a chip would think chiplike thoughts, eschewing the merely human on grounds of aesthetics, or possibly simple revulsion. What's the incentive for thinking like an ape with delusions of grandeur? Sounds like clear grounds for intervention and probable deprogramming.
``Tension, apprehension & dissension have begun!'' - Duffy Wyg&, in Alfred Bester's _The Demolished Man_
There is no doubt that we can already now construct machines that mimic the human brain's functioning and has the same or greater capacity in all areas; but intelligence is perhaps not a good word to use, as it is a very vague term. This is from Wikipedia:
[About intelligence:] A very general mental capability that, among other things, involves the ability to reason, plan, solve problems, think abstractly, comprehend complex ideas, learn quickly and learn from experience.
Things like reasoning, planning and solving problems can all be learned; I'm not even sure how to define "comprehend" in a logically sound manner. Learning and abstraction, on the other hand, are innate to neural networks - one of the common applications of a neural net is feeding a large number of datasets into it, by which process it learns to recognise other datasets that possess the same properties as were common to the learning sample; which in essence describes the process of "abstraction".
Apart from the potential vagueness in the definition of intelligence, isn't it "consciousness" that people actually mean in this context? But we don't really know what that is, yet. "Self-awareness" simply isn't adequate, IMO - all it means is that you are able to distinguish yourself as an object apart from the rest or the world; I find it hard to imagine anything with even a simple brain not being self-aware.
Form (circuitry) does not equal function (brain processes). A situation in which the people involved didn't grasp this fact was the pacific island cargo cults. They built crude replicas of the shiny magical giant birds on the abandoned runways in order to entice the real ones to come back and feed them some more. Extend the circuitry to trillions of switches and quadrillions of connections if you like, it'll still be a lifeless static model and nothing more.
Those billions of neurons are connected by those trillions of synapses in a very important fashion. They are connected so that no neuron is more than six synapses from any other neuron, the average being three synapses. This requires mapping the connections in a very high dimension (a mathematical term, not a throw-away physics term). Look up the Connection Machine for a lesson. In that computer the mere thousands of (RISC to the limit; 1 bit) processors were connected in a 12 D hypercube. The very efficient layout of the 12 D connection circuitry resulted in a box for the CM-1A with 1024 processors being 1.5 meters on a side, or 3.375 cubic meters. 22 billion processors? A layout one machine high would cover 49.5 square kilometers. That's a square 4.35 miles on a side. That's all due to the massive cross-over circuitry needed for the interconnectivity in the design. The circuitry to simulate brain function would require far more volume. I don't suppose you've noticed that they haven't yet done so, have you? I get the feeling that despite thinking they can simulate a brain based simply on circuitry, they don't understand just what sort of circuitry is required. The tip off is the suggestion that the circuitry might be able to be imprinted on a chip. Not even if the chip were the size of the Vehicle Assembly Building at Cape Canaveral, where they put together the space shuttle prior to launch, is my feeling.
I also get the feeling they can't define for us at what point their 'artificial intelligence' surpasses that of human intelligence because they don't have a workable, objective definition for 'intelligence'. They may think they do, but these folks aren't cognitive scientists, and we cognitive scientists know we don't.
They're trying to pack a single kind of circuitry tighter and tighter hoping to simulate a rain. But a brain has many different kinds of circuitry. And that circuitry can interact with other circuitry to either stimulate or inhibit its operation depending on input to the first. That mode selection is done by an entirely different kind of circuity in the switching system of the thalamus, and is given weighting in the form of context according to another different sort of circuitry in the limbic system. They're trying to build up to the equivalent of the neocortex so they can have something like the 'new' part of the brain, but are neglecting the infrastructure of the 'old' brain, which makes the 'new' operate like a brain, period.
They're trying to use processors as computational devices. Those have hard coded calculation capabilities. The brain uses collections of neurons for different calculations, and the same neurons can calculate in different ways according to what collection it happens to belong to at the moment, essentially being a part of a computational collective whose capabilities are reprogrammed on the fly. Also, their processors are switches, which have definite state calculations, like binary. Many brain processes are calculated by continuous 'slow potentials' instead of the binary output 'firing'. Their device can't do this.
They're using that binary output as the basis of coding for transfer of information between processing structures. Brain process output can be coded as changes in firing rate compared to resting 'spontaneous' firing (do they have spontaneous firing as a baseline?), changes in the variance of firing rate integrated over a time scale suitable for a particular process (ie. pain response vs. rational thought), pairs of pulses whose inter-pair and/or intra-pair spacing varies and so inte
"I may be synthetic, but I'm not stupid." -- Bishop 341-B
Two points:
1. It is not necessary to implement a system that simulates the human brain _in real time_, if you can run a simulation which works even hundreds of times slower than a real human then the point is proven. You could then (theoretically) do something as simple as double the numbers of neurons it has, and tell the thing to design a faster version of itself. BUT, see second point:
2. It is not enough to simply simulate the human brain. The mind is nothing without an environment to observe and learn from. All you're going to end up with is the digital equivilent of a brain in a jar, a brain which has less experience to draw from than even a newborn infant. At the very least, we need some way to communicate with the "digital brain". Hooking it up to a camera and microphone, and a speaker for output might be a solution (not possible at present of course, even if we already had the brain); that is if you want a mind that will have nothing in common with a real human and will almost certainly feel alienated from all other forms of life and develop serious psychological issues (try to imagine youself being born into that situation).
The ideal solution of course, is to simulate an entire universe for the mind to interact with, and it would have to be complex enough to evoke feelings of curiosity and the desire to experiment in the simulated mind. You'll never get anything approaching the level of human intelligence unless you provide a world roughly approaching the complexity of our own.
That's requires a lot more computing power than just simulating a few thousand neurons. (Of course, you might be able to simulate the whole thing _really, really_ slowly, see point 1).
Nope, I don't see an artificial brain competing with us anytime soon.
No
Sqreater's law says that a thing cannot make an artifact as complex as itself. That would be "bootstrapping" and is not allowed by this universe any more than it allows you to lift yourself by your physical bootstraps. Our level of complexity is an asymptotic goal that can only be reached by an ever increasing application of effort and resources, like the speed of light. We don't have, and never will have, an infinite amount of resources. Therefore, we never will be able to replicate our own intelligence. And those who project complete health and very long life forget entropy. It's also the law.
E Proelio Veritas.
Well, just keeping with the neural network basics on the assumption they are close enough to how the brain works; or that a neural network is an equivalent type of computer:
The human brain would then be an analog computer functioning with some differences. The network itself and the weights being more important than the cells themselves; well, lets just assume that for the brain; its already the case for the NNet.
I've not read any papers on this, but I have not seen much in the way of transferring knowledge (non-linear approximations) between different networks because the knowledge is in the network and is too complex to map over. The solution commonly used to move the knowledge over is to plug one into the other so the destination re-learns the function. I can't see how a human brain could do this.
Ok, say we have a copy of the brain's network in the computer so we don't have to worry about transferring between different NNets: Then we have issues with the analog nature of the brain. I read about a hardware NNet long ago which used less "neurons" to perform the same task as the software NNet. What was discovered was physical properties were impacting its operations and being capitalized on; therefore requiring less hardware to perform the same task. The downside was the NNet would fail when the outside noise that was part of its learning changed - they changed the temperature and it messed up. I thought this was somewhat telling in that a human brain is much more limited in temperature than the mere living conditions of the cells; it starts malfunctioning too. So, we may simulate the thing perfectly but the analog factors make the two systems too different to simply clone-- so it becomes a problem like I mentioned above. (unless we simulate enough of these factors to get it to "heal" itself-- using a real brain as a boot strap.)
Yes, I think smaller brains will come first. I do think our simple NNet model we have is amazingly powerful and approximate enough a foundation; we do intelligent things with them already and they are extremely small.
I'm willing to bet on such a short time because I think we can figure out enough stuff to get the machine to figure a lot of it out itself. I don't think brains start out wired all the same; they develop with some initial fractal-like pattern which we might be able to model by that time?
Do we need to model a human brain? Can we just get something similar enough that it figures it out itself? Maybe it isn't on par, but if it can approximate us on many tasks won't that be enough? If it does act human, is it? or more importantly, do we really want to make the ultimate rube goldberg machine that just makes the same mistakes we do?
Democracy Now! - uncensored, anti-establishment news
"Biologist P.Z. Myers has criticized Kurzweil's predictions as being based on "New Age spiritualism" rather than science and says that Kurzweil does not understand basic biology. Myers also claims that Kurzweil picks and chooses events that appear to demonstrate his claim of exponential technological increase leading up to a singularity, and ignores events that do not"
Great you've found someone to make this kind of argument but it misses the point. Who care if Kurzweil doesn't understand basic biology? Sure, some fields make vastly more progress than others. The point is, once a particular field makes enough progress, it can simply replace the fields and the methodologies that haven't made progress - a pure silicon "brain" more intelligent than a human being in some respects could certainly provide breakthroughs in many fields where progress hasn't previously made. Another example among a zillion: AI failed to solve chess by simulating reasoning for years but presently, semi-brute-force have simply leap-frogged that challenge and chess programs out-perform humans without appearing to be intelligent.
I think that the pro-singularity argument is that the fastest advancing, strongest part of technology will get us there, leapfrogging the weaker parts. You might contest this but it seem like actually showing why a given problem can't be leapfrogged or gone-around would a necessary ingredient for such an argument. A slowly advancing field proves nothing. What you need to argue for is an inherent roadblock.
No one answered the central thesis of scalability. That was to posit that twenty two of this group's 'superchips' will be able on paper to duplicate an 'average' brain. No one has challenged in reality the essential math of the letter's hypotheses. All these slashdotters and all they can come up with is the tired melange of cynical drunks in a suburban bar. If the letter is really on the level, the idea of duplicating the human brain has been answered right there in the original posting. It is a 'done deal' I used to work on electronics. Anyone ever seen a drawing of an integrated circuit for a memory chip. They will see a large array of identical subcircuits. And so will this first attempt at brain duplication. What is left? Anyone want to download herself or himself into one of these electronic souls? And if a circuit or several 'go crazy', then who is to be consulted? A 'psychiatrist' or a technician? C'mon crew, who wants to be the first of the new Asgard race a'la StarGate SG-1? I just KNOW that Thor is out there somewhere. Or how about Loki? Yeah, that is it, on this site there will be MANY Loki's!
I see no reason that will make constructing an artificial intelligence as complex as our own impossible.
It may take 20 years or 2000, it may be take the form of a digital simulation or a biological "life"; the details don't matter as long as we eventually achieve that goal.
Now, when we have such an entity (manufactured in a fab, grown in a vat, whatever) that passes the Turing test perfectly and is thus indistinguishable from a human intelligence, we will have to contend with the notion that if it isn't different from us, we are not different from it. Soul not required, free will probably too.