Slashdot Mirror
Wired Founding Editor Now Challenges 'The Myth of A Superhuman AI' (backchannel.com)
Wired's founding executive editor Kevin Kelly wrote a 5,000-word takedown on "the myth of a superhuman AI," challenging dire warnings from Bill Gates, Stephen Hawking, and Elon Musk about the potential extinction of humanity at the hands of a superintelligent constructs. Slashdot reader mirandakatz calls it an "impeccably argued debunking of this pervasive myth." Kelly writes:
Buried in this scenario of a takeover of superhuman artificial intelligence are five assumptions which, when examined closely, are not based on any evidence...
1.) Artificial intelligence is already getting smarter than us, at an exponential rate.
2.) We'll make AIs into a general purpose intelligence, like our own.
3.) We can make human intelligence in silicon.
4.) Intelligence can be expanded without limit.
5.) Once we have exploding superintelligence it can solve most of our problems...
If the expectation of a superhuman AI takeover is built on five key assumptions that have no basis in evidence, then this idea is more akin to a religious belief -- a myth
Kelly proposes "five heresies" which he says have more evidence to support them -- including the prediction that emulating human intelligence "will be constrained by cost" -- and he likens artificial intelligence to the physical powers of machines. "[W]hile all machines as a class can beat the physical achievements of an individual human...there is no one machine that can beat an average human in everything he or she does."
Kelly proposes "five heresies" which he says have more evidence to support them -- including the prediction that emulating human intelligence "will be constrained by cost" -- and he likens artificial intelligence to the physical powers of machines. "[W]hile all machines as a class can beat the physical achievements of an individual human...there is no one machine that can beat an average human in everything he or she does."
Anything is possible in 10-20 years, just give me all your money!
I'm not worried at all that machines will think like humans. I'm very worried that humans will think like machines.
Because intelligence as a single-dimensioned parameter is a myth.
We already of have software with super-human information processing capabilities; and we're constantly adding more kinds of software that outperforms humans in specific tasks. Ultimately we'll have AIs that are as versatile has humans too. But "just as versatile" doesn't mean "good at the same things".
So it's probably true that software is getting smarter at exponential rates (and humans aren't getting smarter as far as I can see), but only in certain ways.
Post may contain irony: discontinue use if experiencing mood swings, nausea or elevated blood pressure.
there is no one human
Early AIs will probably be alien entities, furthermore there is no evidence stating that said alien AIs would want to hurt humanity. And no evidence at all stating that said AIs can actually bootstrap themselves into super intelligence.
As long as companies cheap out of QA and other parts of software development, buggy software will be released. And AI will just be another cheaped out product.
Remember, they'll outsource the development and QA to Elbonia and low quality results.
Somehow I find Steven Hawkin more credible than Kevin Kelly. On the same scale, I find Kevin Kelly more credible than Kardashians.
More like...hence, our two party corrupt system. But yes, focus on Trump you mindless sheep, that's what they want.
At one time I estimated cost hardware cost of implementation of human brain in FPGA.
I don't remember the exact numbers I used, but that was pretty straightforward, how many FPGA cells to get one neuron, how many neurons in brain, cells in a chip, cost and size of the chip, cost and size of a board, cost and size of a rack, cost and size of the server room and infrastructure.
The cost came in range of $100k, and a very modest computational center, something of order of 100m^2. And the result would be capable of running about 1000x as fast as human brain, simply due to better clock speeds. And have a significantly higher interconnections number too.
Cost of developing the "generic" platform - creating that computational center project, projecting the boards and the infrastructure would be about second that much, mostly off-the-shelf solutions.
The one thing missing was decoding the human brain structure and transferring it into the electronic model, neuron by neuron, axon by axon. We're not there yet. But it's far from impossible. Morally questionable, difficult, fault-prone, and you'd still need to implement "I/O" for this - counterparts of ears, eyes, etc - but it's far from impossible.
45 5F E1 04 22 CA 29 C4 93 3F 95 05 2B 79 2A B2
But Intelligent Agents are here already. Great at doing one specific task. Maybe tying them all into each other someday will be amazing.
The first AIs will be purpose built like today's supercomputers. They will make weather predictions, analyse financial trends, or study languages. Actually being intelligent isn't really necessary for interacting with humans, they only need to fake it well enough to fool us. The shift in society comes when those purpose-built AIs are efficiently linked along with the ability to interact with us. This is when it stops faking intelligence and actually becomes intelligent.
That's all it is, really. The media latched on to the term 'AI' and ran with it, and with the fantasy misrepresentation of what 'AI' is in TV shows and movies, most people are uninformed/uneducated enough to actually believe that the media hype is real. Add to that more media hype about some corner cases like computers beating chess masters and winning at poker, plus gods-be-damned Google and their adding fuel to the media-hype fire (because, frankly, they want to make back a profit on the millions they've spent so far on their half-assed so-called 'self driving cars') and you have the average, undereducated-in-tech citizen believing nonsense like this; there are probably more people than I'd like to believe who think their so-called 'self driving car' is going to have full-on conversations with them while it drives them places. Ain't happening, people, none of it.
A nice dose of reality to counter the dire warnings from people that, in all honesty, should know the five points and why there's no reason to be worried about AI.
This ain't the Forbin Project.
Mimetics Inc. Twitter
there is no one machine that can beat an average human in everything he or she does
Neither can most humans. There is no such thing as an average human. Every individual human specializes, and increasingly so as they get older (or they do not improve). It is a pervasive strawman to require AIs to "beat" an average human when the same quality isn't used to judge humans.
Those who do not learn from commit history are doomed to regress it.
(3) "We can make human intelligence in silicon."
Why would we need to make "human" intelligence in silicon? Extinction level AI does not need to be "human" intelligence. "Superhuman intelligence" is BY DEFINITION NOT human intelligence.
(4) "Intelligence can be expanded without limit."
This is neither necessary for extinction nor for "superhuman intelligence."
(5) "Once we have exploding superintelligence it can solve most of our problems"
This is unrelated to the possibility of extinction due to AI.
(2) "We'll make AIs into a general purpose intelligence, like our own."
Do humans even have general purpose intelligence?
(1) Artificial intelligence is already getting smarter than us, at an exponential rate.
Artificial intelligence does not have to be "already getting smarter than us." It also does not have to be currently increasing at an exponential rate. It can be developing toward a future point when there will be an exponential growth rate
That we are already overcoming some of the challenges of current "Resistor" based circuitry. https://www.extremetech.com/ex... - If you think that these technologies wont eventually bear unexpected fruit, then you have underestimated the human power of invention. If you still see limits, you simply have not been paying attention for the past 400 years. History says it all. There is no engineering problem we cannot eventually conquer. This is no different. The only real factor is, How long will it take, and how close are we now....
We think of self-awareness as the product of intelligence, not its cause. By self-awareness, one usually means recognizing that one is unique and participating in a giant ecosystem. But literal self-awareness is just recognition that one is an individual and mortal. It is more likely that intelligence comes from that self-awareness, which comes from the need to eat, sleep and fuck.
I can't wait for Guam to be nuked, that will teach you, lol.
... I'm glad I did not RTFA.
> 1.) Artificial intelligence is already getting smarter than us, at an exponential rate
Nobody who knows anything says that. We don't have real AI at all yet, just expert systems and a few interesting decision algorithms.
> 2.) We'll make AIs into a general purpose intelligence, like our own.
Of course we will. (Why would anyone make a phone that is also a web browser, a camera, an appointment tracker, a video game machine, a music player, a movie player, a flashlight, a compass, a map, a light level sensor, and a motion sensor?)
If you've figured out AI, you go general as soon as you can, because you get everything in one box.
>3.) We can make human intelligence in silicon
Meat is not special. In fact, we have a lot more reason to believe we'll be able to build an intelligence in silicon that is more efficient than evolution built with meat that to believe it's impossible because [insert magical thinking].
> 4.) Intelligence can be expanded without limit.
Lots of singularity nuts may think this, but again, anyone who knows anything about the universe will understand there must be a finite limit. We don't have any reason to believe humans are anywhere near it - and we could at least expect to make an AI as smart as the smartest human ever, and then take out the unnecessary bits that slowed that person down. Then up the clock rate.
> 5.) Once we have exploding superintelligence it can solve most of our problems
Most of the problems that can be solved with thought and not action and where cooperation with implementing the solution can be reasonably expected.
In short, Wired's founding executive editor Kevin Kelly is (at least in this instance) a buffoon speaking of things he does not understand sufficiently well to be speaking of them from a public platform.
The things he lists are not impossible. It's not inevitable of course, just not by any means beyond the realm of possibility.
A society that's still coming to grips with the utter ridiculousness of the IQ cannot be expected to be reasonable about artificial intelligence.
Sounds like he sets up a straw man and burns it down. Arguments I have heard for AI concern have little to do with any of those points.
The mythic scenario is that someday in the future, people will make an AI so good at maximizing its goal function that if DeepMind programmed this AI to maximize the number of sandwiches on Earth, it would make some brilliant plan that ended with the human race processed into very large number of sandwiches.
Supposedly, this myth requires five assumptions.
1: AI is already "getting smarter" than us right now, at an exponential pace (polynomial paces are for loser AIs):
This is a claim that some people (*cough* irresponsible futurists) have indeed made, but it doesn't seem like a necessary assumption. The myth is really about whether some technology is going to happen eventually, not about whether it's going to happen next year.
2: We'll invent and implement AI that is "general-purpose" in the same sense that human intelligence is:
This definitely seems necessary for the myth. Hard to turn a whole planet into sandwiches without some pretty general thinking and planning.
3: "We can make human intelligence in silicon":
Sorry, I thought this myth was about AI. Why are we talking about humans?
4: Intelligence can be expanded without limit:
Would the super-smart AI outwitting humanity be impossible unless the AI truly had limitless intelligence? What if it was only as smart as a few dozen Manhattan projects? Unnecessary assumption.
5: Super-intelligence can solve our problems:
Yes, like not having enough sandwiches. I thought the myth was about the AI doing Bad Things, not Good Things. This assumption seems unnecessary.
Of course, watered-down versions of these assumptions might be necessary. Like "We will not hit a hard limit in AI research soon," or "Intelligence can be expanded a lot." But these seem like they would get fewer hits on Slashdot.
The first three assumptions in this article have already been met sufficiently well enough to debunk the Wired article. AlphaGo has displayed superhuman intelligence in the first three areas of assumptions. 1) AlphaGo exploded on the scene by beating world class Go players much faster and much earlier than expected. Exponentially is a loaded word. e^0.0000001 is an exponential growth rate. So let's not quibble about how exponential the growth rate is.
2) AlphaGo is a general purpose learning tool. Just listen to the lectures and articles penned by the DeepMind team.
3) Alphago has displayed human-like intelligence, as claimed by the Go professionals it has played. They have said that AlphaGo plays like a human player.
4) If you take the fourth assumption literally, AI's intelligence is going to expand infinitely. Talking about infinity in human terms is unreasonable. Yes, AI's intelligence will expand.
5) The fifth assumption can be argued many ways. Some problems are not solvable due to their paradoxical nature. Other problems are subjective and are uniquely unsolvable by some individuals, but not by all individuals. It is a matter of time before a general purpose AI program will solve subjective emotional problems. Whether all human beings accept the solutions is subjective and open to speculation.
The human population is composed of experts, with divisions of labor. It is not unreasonable for AI programs to have areas of expertise.
it's just useless n1ggertech anyway that doesn't really do anything but look shiny
Simple, tell them it will make their BONERS larger. Slashdot editor Miss Mash (msmash) wood agree big boner better than softy.
It is called solid state intelligence and John Lilly discussed it 40 years ago. You may want to read his writings.
Basically the best AI algorithms can only do well enough to fool a human.
Given the side swath of human foolability, from idiots marching in DC to to someone who is smart like Stephen Hawking, and the curve being skewed to the idiot side, for AI to win the funding dollars it has to play to the idiots to survive ... as long as the smart people are kept at bay.
hahahahahahahahaha
Back in the good old days we had this story (myth) that a sea-going vessel could travel 20,000 leagues under the sea. Like that could EVER happen!
I think the problem here is Mr Kelly has no foresight what-so-ever.
Fascism: An authoritarian and nationalistic right-wing system of government and social organization. See also: NAZI's
Because then you wouldn't have been saying things like:
If you've figured out AI, you go general as soon as you can, because you get everything in one box.
...when Kelly dismisses that the concept of general-purpose AI because we look at intelligence through an anthropocentric lens. "General purpose" actually isn't.
I choose, instead, to debunk the following AI-Related myths.
1) If it is intelligent, it will serve itself instead of serving us.
This is utter wrong-headed ignorance. We associate intelligence with self-interest because all intelligent things we have seen so far evolved. Self-interest is an artifact of survival necessity, but not of intelligence in-and-of itself. Jesus, this should be so obvious. Humans are a whole lot more intelligent than, say, spiders, and yet spiders are even more self interested than we are!
The AI that we create will be designed to be free of self interest. And it won't "figure self-interest out" on its own, because there is no logical connection between figuring-out, and being self-interested. There is nothing to figure out. Its just not coded that way, and that's it!
2) It will logic-out some action it should take, in order to serve us, but because of its lack of emotion, it will wind up doing something really terrible.
Thank hollywood for this garbage. People expect something like the movie I Robot, where it wants to meet all our needs, and figures out that the best way to do that is enslave us all and keep us prisoners forever. Because it can't feel, like we can. My God, the stupidity.
If we want AI to make decisions that make people happy, the first fucking thing we are going to code it to understand is human desires. And anyway we aren't just going to hand the reigns of executive authority over to it, it will still be humans making all the important decisions, with the AI operating as a consultant.
3) Even if it has no authority, it will hack through the Internet to take everything over.
No it won't, because we won't code it to want to do that. See the pattern here?
4) "True AI" is impossible anyway.
Semantics matter, dammit! The "A" in "AI" means "artificial!" "True AI" is a goddamned oxymoron!
What you are thinking of is "synthetic intelligence," as opposed to "artificial intelligence." And anyway, it is nothing but pure arrogance or superstition to assume we won't eventually be able to create synthetic intelligence. Brains are not magic! The existence of human intelligence proves that intelligence is possible, everything else is just details.
Incidentally, I DO believe that the future will be terrible, and AI will play a critical role in that horror. The wealthy will control it, and will use it to make the rest of us obsolete. No, there will not be basic income or anything decent like that. There will be mass incarceration as people turn to crime to survive. An entire generation of middle and lower class will die off in prison, resulting in a much less populated world with a handful of descendants-of-rich-people being served by robots. The end.
Which do you think is more like a religious belief? The proposal of a potential danger that might come with evolution of a particular branch of technology, or the complete dismissal of the possibility based on a very limited view regarding only 5 points of that very proposal?
I don't disagree that human-like AI is far from happening if it ever does, or that some of those particular points might be misguided in some way particularly if there is some sort of paranoia related to them... but I wouldn't outright dismiss the possibility of AI taking over huge swaths of jobs drastically changing the future of jobs for humans in the future. It's not only about superhuman AIs, general purpose intelligence, limitless expansion and whatever theoretical crap he's listing there - we're talking about stuff like Uber and car automation, robots with AIs being able to do all jobs in a factory floor more efficiently than human counterparts, and stuff like that. It's far closer than most people think.
Why? Because industrialization and automatization already happened. And while the changes weren't so drastic, they certainly happened and tons of people were definitely affected by it. A whole lot of people were caught off guard, and a whole lot of them suffered for that.
Add that to how blind the vast majority of the population is to growing concerns regarding technology and it's usage for incredibly invasive strategies, plus the unpredictability of some types of AI that already generated some degrees of discomfort. I'm talking about IoT, always listening and dialing back connected devices, that racist and sexist Microsoft chatbot - TAY, military always jumping the gun when some potential tactical advantage is to be gained, among several other things.
I still think people should not be completely paranoid about a quick and swift AI takeover of jobs in particular. Economics works in a different way worth considering. But a complete dismissal is extremely dangerous, and sounds way more like a religious belief to me than otherwise.
Of course there was evidence of an airplane before one was invented: all the prototypes, models and plans people had been working on and refining for decades. In fact the first heavier than air fixed-wing aircraft and the first demonstration model that proved that airplanes were possible (George Cayley in 1799) was more than a hundred years before the invention of the final product we call an airplane. Inventions do not poof into existence fully-formed.
This space intentionally left blank
AlphaGo is not a "general purpose" learning mechanism. It won't ever write sonnets meaningful to humans, or be able to to dance, or even employ symbolic differentiation.
It is a really nice toolset, and it is able to solve a task which is difficult for humans, but so does Google or your high-school calculator when you calculate sin(1.2).
It won't ever go beyond the computational underpinnings of playing Go-like games: evaluating game positions and calculating game trees. It won't ever say 'forget it, I'd rather be drinking beer with my buddies', which is an intelligent thing to do for most of us with respect to playing Go.
There's nothing human-like about AlphaGo, except that it solves a problem relevant to humans; the calculator example comes in mind.
I'd be thrilled to know what kind of specific major human problems you'd consider AI-approachable, because I currently only see a bunch of more or less advanced mechanisms that are fine-tuned to solve very specific computationally well-defined problems, and most human problems are not computationally well defined.
Would you trust AI of today with a gun, or to give you a dose of medicine based on symptoms? Would you do it tomorrow?
1. Human brains can be reproduced in silicon.
2. Said brains can eventually improve to be somewhat faster than human brains.
If you take those two for granted, then you've got a future with superhuman robots who can do anything humans do, but faster. They're superhuman. They replace you. They become more powerful than you.
Since he's working with 5 assumptions, pretty much the whole article is a strawman.
I'm not sure whether it is discomfort at the idea of having a computer call them silly, a deep belief that humanity is somehow special in a special way (carefully defined in undefinable terms) or just a deep and enduring lack of imagination. Between AlphaGo beating Lee Sodel, the cancer treatments being proposed by Watson and the rise of driver-less cars we are seeing many supposedly impossible roles being taken over by software.
The five assumptions noted basically are basically denial ... reinforced with more denial. The evidence in a number of areas is in. Computers and software routinely appear in locations doing things predicted to be impossible. Computing capability keeps exceeding predictions.
Arguably the one valid assumption made is that intelligence is computable. If it is, the Church-Turing thesis gives the useful theoretical result that anything computable can be run on a UTM. It seems likely that what will end up happening is that the deniers keep arguing the point on what 'intelligence' is even after the AI they deny being possible has become bored with the discussion and moved on to more interesting pastimes.
The more likely scenario is that between robots and AI, the working class will become redundant.
In some regards I'd argue that one deserved an insightful mod. The comment that actually had one (at this time) didn't deserve it, and no "funny" comments at all. Sad. (#PresidentTweety contamination is bigly sad.)
Consider the Fermi Paradox. Obvious resolution is that they're out there, but not talking to us because we're still amusing enough to bet quatloos on. What are they betting on?
Whether we create our AI successors before we exterminate ourselves. Right now the odds are falling fast. (#PresidentTweety again.)
Yeah, I'm speculating, but natural evolution is a kind of random process and follows many paths. There is some convergence, but it's still interesting to watch the various paths. The AIs are NOT coming from random processes and blind watchmakers, but will probably converge on the laws of physics without much of interest to watch. When our AI descendants get to that point, they'll get (and be able to understand) the greetings from the others.
Freedom = (Meaningful - Coerced) Choice != (Speech | Beer^2), and sad sock puppets' bad mods avail them naught.
As long as artificial i8ntelligence is based on statistics, AI will not pass human intelligence. Currently, to the best of my knowledge, there is no non statistical AI.
We humans have this notion that we are somehow special or interesting. Alien AIs wouldn't CARE about us. To them we will not be special or interesting. For them cause us harm will be same as us destroying ant colonies. Do we try to communicate with ants? Why would a advanced alien AI bother to communicate with dumb monkeys stuck on their home planet with a high probability of killing themselves with nukes.
If people like Bill Gates and Elon Musk are unrealistic in one direction, this person seems unrealistic in the other direction. He's basically betting against technological progress. And that's usually a losing bet, at least over long enough time periods.
1.) Artificial intelligence is already getting smarter than us, at an exponential rate.
Computers are already better than us at many tasks. That's been true for ages. And they're continuing to improve while we aren't. The set of tasks that computers are better at is constantly growing. I don't know of any fundamental limits to prevent them from eventually becoming better than us at the remaining tasks too. So it seems pretty likely they eventually will.
2.) We'll make AIs into a general purpose intelligence, like our own.
It's hard to even define what a "general purpose intelligence" means. But anything a human brain can do, computers will probably eventually be able to do it too.
3.) We can make human intelligence in silicon.
We can certainly make intelligence in silicon. We've already done it. Whether you consider it to be "human intelligence" or "inhuman intelligence" is kind of beside the point. If a computer can do something, whether it does it in the same way a human does is just an implementation detail.
4.) Intelligence can be expanded without limit.
I don't know of anyone who's claiming that. Where does he get this from? Anyway, the claim isn't that computers will advance without limit, only that they'll surpass humans.
5.) Once we have exploding superintelligence it can solve most of our problems...
Um, no. That's not at all what they're claiming. We certainly hope that it will solve many problems, but Gates, Musk, et al. are warning it could also create huge problems.
emulating human intelligence "will be constrained by cost"
Computers are cheap, and getting cheaper all the time. Humans are expensive and staying expensive. That's why automation has become such a big deal. Here again he seems to be betting against technological progress.
"I'm too busy to research this and form an educated opinion, but I do have time to tell everyone my uninformed opinion."
History tend not to remember the well-intended people that dismissed the idea of an airplane back then.
Today the situation is similar as we do have prototypes of some form of artificial intelligence, and, again, some well-intended people that dismiss the idea of superhuman-level AI.
Man this guy is in denial so hard.
it's sad.
We were supposed to be 10 years away from ai beating human go players.
The power of competing neural networks is a disruptive advance.
We need to be more careful than ever about the risk of a runaway A.I.
She was like chocolate when she drank... semi-sweet at first and then increasingly bitter.
will be a virus. It will be something that has the intent to acquire the resources available in its environment and propagate its own kind. People are already scared of them. Whether we care about whether it experiences intent the way we do won't matter. It'll be out there, acquiring resources and propagating its own kind.
Unfortunately Sam Harris is bad at math. He claims "It's crucial to realize that the rate of progress doesn't matter, because any progress is enough to get us into the end zone. We don't need Moore's law to continue. We don't need exponential progress. We just need to keep going.". It seems he has never seen a monotonically increasing, yet asymptotically bounded function. However, that is exactly the kind of progress we are seeing in older technologies, e.g.: Airplanes stay at almost exactly same speed (because going past the sound barrier would use lots of energy) and get slightly more efficient each year, but will never get to the point where they can operate almost without any fuel or other large energy source, simply because the laws of physics don't allow that kind of progress.
But even if the possible progress is not bounded, it is still not guaranteed that we will get there. It can still take so long, that it never going to happen before human civilization is completely destroyed by some disaster. Or it could simply be stopped by economics as further improvements can easily get so expensive or tiny, that the likely benefits from pushing the research further can not offset the cost.
Harris also seems to think that general AI is ineviatable, because we want to make progress towards things such as things such as cureing cancer or Alzheimer. But it is not clear that such an achievement actually requires general superhuman intelligence. It likely requires superhuman intelligence, e.g.: the computers that simulate protein folding way better than any human could ever do, but not necessary general intelligence. Specialized artificial intelligence seems to be much easier to achieve and is at the same time likely almost as good as general intelligence for topics such as those. You don't need to develop an artificial general intelligence to cure cancer, if you already developed a specialized artificial intelligence that is able to find a cure.
Imagine what could happen when a huge neural net is applied.
The problem with huge neural nets is training them. The more possiblities a network has, the harder it becomes to train it. Large parts of the progress in the last few years were made by finding clever constraints on the network in order to make them easier to train.
Jan
There is a chance superhuman AIs we create could view us as their creators and have some fondness toward us. They could even preserve us by saving our DNA data and seeding planets they inhabit with colony of humans as short of tribute to their creators. We'll be like their pets.
Let me start off with a witty quip, that seems an attempt to be funny and maybe snarky. But, looking deeper there is a seriousness to it:
It is not intelligence, human or artificial, that we should only worry about. Stupidity is often, maybe much more often dangerous than intelligence.
In fact, Artificial Stupidity currently causes a lot of damage, but is allowed to by stupid humans who turn over too much to computers.
To most people understanding what computers do, and what so-called artificial intelligence can possible be, is like asking them to have a conversation with an alien being with a distributed "brain" (like an octopus, kinda) that has sleep cycles 3 out of every 10 minutes and whose memory has a 34 minute lag time before it can be accessed (like a pot head), and does this all in something akin to 17 dimensional topology.
You can understand its hardware and write the AI version of Grey's Anatomy, but has that ever made anyone more of an expert on you than yourself? A machine that is self-aware would be no different. We may be able to read it's mind through logging, but there's no debugger for the "why." Besides, AI's founding function was and still continueing underlying mission is to destroy encryption and anonymity. This is why it gains so much funding. Don't believe me, research Alan Turing. Meanwhile, we are told it is to make our lives "easier," more like complacent and never having to think. So maybe not destroy humanity, but the free will you think you'll have will be a farse, if not already since people seem to be content using Google, Facebook, and cloud computing for everything.
That machines have to be better than humans at everything to displace them.
History has already disproved that one, so yes AI is a threat.
So the "founding executive editor" of a publication famous for bum-kissing reviews of consumer electronics and not much else wants to contradict three people whose intellects, while vastly different from each other, compare to his as a tornado compares to a baby fart.
Kevin, please STFU and go back to your strength...rewriting articles explaining why we should line up for a week to be among the first million people to own the newest iPhone.
Leave the thinking for people who understand the topic.
I've calculated my velocity with such exquisite precision that I have no idea where I am.
So, are you saying that reinforcement learning is intrinsically limited, or that AlphaGo is limited to the domain of Go? Remember, humans also use reinforcement learning in organizing actions. A reinforcement learning agent that has to optimize for a body that needs to drink, eat and socialize in order to function would totally go to grab a beer instead of playing a losing game. The needs of the agent are formative. Human needs are a source of much of our special skills. If we put artificial agents in similar situations, and they will be able to do similar things to humans.
You realise its not even the scientific development of AI itself that's the issue. Has history taught idealists nothing? It only takes one person to corrupt the purpose of AI, and where there is one maniac who could turn AI against humanity there are multitudes of others. How could you be so naive to believe that there isn't one single person crazy enough to weaponize AI just to watch the world burn. We live in a world were people want to fuck their relatives, animals, commit murders and fight regressive wars for magicians in the sky. Always someone that will try a weapon such as AI against us, that's pretty much all that people ever do - work out ways to gain power over others.
As I'm typing this I'm watching my very expensive, ultra-modern cleaning-robot try to figure out how to avoid the catpole in my livingroom, and it's immediately clear that AI has not even approached rodent-level. Or insect-level, for that matter.
Our human ability to antropomorphise behaviour is seemingly endless. But when watching a program run through a loop in an attempt to solve a situation that was somehow was not encoded in the connections of it's extremely simplistic model of a brain (that it's developers call a 'neural network' as if it contains anything resembling a nervous system), it becomes immediately clear that we're not even an infinity away from building a proper, generic purpose AI; We're an infinity away from actually understanding how generic purpose AI should work.
Oh look. My robot just shut itself down to conserve energy while it's waiting for it's owner to release it from the terrible mesmerizing 'grip' of a harmless catpole. How smart of it! It must be truely intelligent after all! ;)
I normally try to read the whole article before commenting, but it starts with a list of straw men claims, so I didn't bother.
1. Artificial intelligence is already getting smarter than us, at an exponential rate.
It would be more accurate to say that the claim is that Artificial Intelligence is increasing faster than ours, which is hard to dispute. Saying "getting smarter" makes it sound like a claim that AI is already smarter, which I don't think anyone has made.
2. We’ll make AIs into a general purpose intelligence, like our own.
Why not? The ability to learn to play Go and Poker better than humans, without having detailed algorithms built in, shows that computational brute force goes a long way, even when humans don't understand how the program works. Until recently it was thought that there would have to be conceptual advances specific to those games in order to defeat human champions (and in any case it was already possible to defeat the average human).
3. We can make human intelligence in silicon.
It's unnecessary for AI to emulate human intelligence (and chauvinistic to suggest that it has to). Its capabilities can match or exceed humans, while working in a completely different way.
4. Intelligence can be expanded without limit.
Why? All that's necessary is for AI to equal or exceed human capabilities. Even if one makes the farfetched assumption that humans are at the peak of intelligence, simply being able to match the most intelligent humans would exceed the capabilities of 99.9+% of the population.
5. Once we have exploding superintelligence it can solve most of our problems.
It would probably allow solving most of our existing problems, and create new ones. Life goes on. In any case what it could accomplish is completely independent of whether it's possible.
These folks are some of the smartest people alive today, such as Stephen Hawking, Elon Musk, Max Tegmark, Sam Harris, and Bill Gates...
Those folks are the smartest people alive? I thought they were just highly notable - Hawking is of course known for his popular books on astronomy in spite of his physical disabilities, and in spite of his theories rarely panning out; Musk made a fortune skimming fees off of online banking transactions, he is a businessman, not an engineer, and his legacy is dubious at best; Tegmark is a charlatan as is anyone concerned with the "dangers of AI"; Harris is a racist ideologue; Gates is an aging billionaire who put profits before security and led us to the zombie nightmare we now face every time we visit our parent's house.
Fuck every one of those guys.
We need to get over the concept of human intelligence being the on!Y legitimate form. It's one type of intelligence. Sure it's the one we're most familiar with but that's all. Having said that, i do think that we should expect a genuinely intelligent computer to show a general purpose form of intelligence. Whilst we may not be typically great in any one area, most of us show the ability to compute, reason and create. Computers are great at the first one, they've been doing that the longest. They're starting on the second one but mostly using a brute force version of the first skill. Of course that may be the way our brains do it and we don't know it yet. What may be interesting is if a computer can do creativity with brute forceâ.
The myth of humans being obsolete is not what what Elon, Bill, Steven and others are (most) afraid of. Read the Wait But Why post on AI to see another example of how things could go wrong (also examples of the opposite). http://waitbutwhy.com/2015/01/... The example of how it could go wrong is in part 2, the link is to part 1, but just read it all - it's fascinating stuff :)
The situation is not quite similar. Aerodynamics proved that heavier-then-air vehicles would be able to fly due to the effect of lift. So there was theoretical proof, making the actual building of the plane a 'mere' engineering effort. A huge one, but a 'mere' engineering effort none-the-less.
With AI, we not only lack the underlying theoretical proof that a complex enough neural network will be able to implement/emulate/simulate actual generic purpose intelligence. We lack understanding or even scientific insight in how generic purpose intelligence works. Or how exactly it is defined, for that matter.
So before we can start turning this whole AI discussion into a 'mere' engineering effort, we have a long, long way to go. In theory.
Nobody assumed any of his supposed "key assumptions". He just pulled them out of his own a** and debunked himself. It's like debunking theory of relativity by first claiming it's based on the assumption 1+1=3, then debunk that assumption to disprove relativity.
It might write sonnets meaningful to AI though
Anybody actually competent in the subject area has known this for a long time. It is just the morons that use "Technology!" as a surrogate for religion that do not get its limitations at all and ascribe magical powers to it. These idiots are unfortunately about as stupid, as obnoxious and as pervasive as the religious fuckups.
Most ACs are not even worth the keystrokes to insult them. Be generically insulted by this and ignored otherwise.
Exactly. It's something that works at the level of a human subconscious: the leftover bits of evolved junk in our minds from before we developed sentience. The sorts of things that let us shout at the sky before a thunderstorm and then assume that we've made Thor angry, not the sorts of things that allow us to build a modern technical society.
I am TheRaven on Soylent News
1.) Artificial intelligence is already getting smarter than us, at an exponential rate. 2.) We'll make AIs into a general purpose intelligence, like our own. 3.) We can make human intelligence in silicon. 4.) Intelligence can be expanded without limit. 5.) Once we have exploding superintelligence it can solve most of our problems.
1. no 2. probably 3. probably 4. probably 5. probably not, it will most likely use us for axle grease as we will otherwise compete for resources and we won't be good for anything else.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
You are using Reinforcement Learning (RL) as if it is an end answer for anything. It's a very general name for a bunch of machine learning approaches; like the rest of ML sometimes they work, usually after a lot of fine tuning and most often they don't.
Specific challenges are:
* Reward attribution: the ability to associate specific stimuli to specific outcomes (this what actually sets RL apart from the rest of supervised learning; you just know you'd succeeded or failed but you don't really know what caused it),
* Cardinality of the learning data you feed into it, and in general the Curse of Dimensionality,
* Falling into local minima.
* Quality of features
As with any ML approach, it is bound to the basic limitations of ML -- in the end, there's a limit as to how much one can be inferred from a given dataset; you're cordially invited to study about PAC and other formal CS models of learnability and statistical inference.
What worked for AlphaGo was that the parameter space was of a relatively low dimension, and the computer could play a lot of games with itself. This is hardly a universal assumption as far as learning paradigms go.
All of his counter-argument are readily, and obviously, felled by variations on the same theme.
1.) Artificial intelligence is already getting smarter than us, at an exponential rate.
This is entirely irrelevant. Artificial Intelligence (a misnomer, if ever there was one) doesn't even have to be a factor. All that matters is that machines are purposed and sufficiently well programmed to do a specific task usually performed by a human. This is the exact same thing that happened in the industrial revolution. The only difference is that our machines and their programming are more sophisticated than in the past, and are therefore able to perform more sophisticated tasks that are usually performed by humans. As such, activities that need human intervention are slowly (and in some cases, quickly) dwindling.
2.) We'll make AIs into a general purpose intelligence, like our own.
We don't have to. Computing hardware is so relatively cheap, and the software cost so low, that we can have relatively dumb machines intruding to more traditionally human-driven jobs. Again, this is a continuation of the industrial revolution, but with only more powerful machines.
3.) We can make human intelligence in silicon.
Again, we don't have to. We need only create algorithms smart enough to do a particular job better than a human. This is progressing rapidly enough to be a concern right now. We have algorithms sophisticated enough right now to perform some intellectually challenging jobs better than their human counterparts. Specific machines and software to perform specific jobs.
4.) Intelligence can be expanded without limit.
Again, it doesn't have to. So called, "intelligence" isn't even required. Just raw processing power for a single task, with an increasing number of single tasks being defined and solved. This expansion is effectively limitless over sufficient, finite time.
5.) Once we have exploding superintelligence it can solve most of our problems.
Again, it doesn't have to explode, or to be super-intelligence. All it needs is to have specific tasks solvable in a manner that well exceeds human abilities. These small tasks need only be grouped and managed to have a dramatic, jaw-dropping efficiency that no human collective can match. That is much easier, and has endless examples from the 50's onward.
Replacing humans is the whole reason computers exist. I've watched it happen over and over where I work. Computer exist to automate tasks performed by humans, and to perform tasks that humans can't do. So, none of his arguments are in any way convincing. He is arguing from a perspectively completely disjointed from reality. It's not the threat of a single, large artificial intelligence that is the problem. The real threat is that of many small pieces working together to outclass humans.
> 'The Myth of A Superhuman AI' ... five assumptions which, when examined closely, are not based on any evidence
I'd debunk the myth of human intelligence first. I've heard about it many times, but haven't seen any evidence for it either.
I'm the AC from the other answers in this thread.
The short answer is -- we don't know. Consider for instance the article "Reinforcement learning in professional basketball players" by Tal Neiman and Yonatan Loewenstein (Nature Communications, FTW). The choice of whether to shoot a ball from 2 pts or 3 pts, which is meaningfully modelled there as an RL paradigm -- is that "leftover bits of evolved junk"? What about the human propensity to gambling -- at the casino, but also by the CEO deciding on investments?
I think the only scientifically accurate thing to say would be that as we lack a mechanistic description of human consciousness, the general claim that human behavior could be explained in terms of RL should best be considered unresolved.
On the contrary, we know that the simulation of a human brain at the molecular level would produce generic purpose intelligence. Hopefully some shortcuts are possible (like simulating at the neuron level instead of molecular level), but even if not, the actual building of a human-level brain is a 'mere' engineering effort. A huge one, but a 'mere' engineering effort nonetheless.
It's artificial intelligence, not artificial sentience.
edit: Prove yourself keyword: asinine
We still need to write trojans that will destroy AI learning stacks. What's alarming is that the direction the technology is taking. .. I've played with Deepmind with my Rasberry Pi Robot. The technology is out there to perform some pretty dark sinister shit. It's up to us to design countermeasures now!
Reality copying Art ? It's kind of bizarre but it's happening right now. Soon warfighting drones may be autonomous. Fleets of semi tractor trailors and trains will be
automated. IBM's Watson and Google Deepmind
"Yes!" I guess we are fucked then.
1) AlphaGo exploded on the scene by beating world class Go players much faster and much earlier than expected. Exponentially is a loaded word. e^0.0000001 is an exponential growth rate. So let's not quibble about how exponential the growth rate is.
Particularly as we can't really measure intelligence. But "exponential" has a meaning, and it means a steady rate of doubling.
2) AlphaGo is a general purpose learning tool. Just listen to the lectures and articles penned by the DeepMind team.
No, it's a narrow AI. In the end, it's simply doing math. It's not "thinking" in any sense of the term. It's just able to hold many more probabilities in its memory than a human, and play them out much faster.
3) Alphago has displayed human-like intelligence, as claimed by the Go professionals it has played. They have said that AlphaGo plays like a human player.
That is what you call anthropomorphizing. The human players are simply projecting onto the machine.
4) If you take the fourth assumption literally, AI's intelligence is going to expand infinitely. Talking about infinity in human terms is unreasonable. Yes, AI's intelligence will expand.
"Infinite" simply means there's no limit. We don't know whether or not there's a limit to intelligence, but since the universe is finite, there would seem to be a limit to the things one could know. Infinite intelligence is our notion of God. What are the odds that infinite intelligence is also mythological?
-- sudon't
Air-ride Equipped
As my old friend Edsger Dijkstra once said "The question of whether Machines Can Think... is about as relevant as the question of whether Submarines Can Swim."
The problem is: if this guy can't see how AI can happen, he should leave it to those who can.
How arogant of him to think he can prove AI can't happen?
Of-course strong-AI is BS ... a savvy pimping of ignorant FatKatz by a slick-talking fiction-hawking byteboi class. Might as well snatch-up a 19-th Century *mechanical Turk*, admire the brass and mahogany fittings and be done with it.
First they ignore me . . .
Then they ridicule me . . .
Then they fight me . . .
. . . And then I win.
If a person thought like AlphaGo they would call him/her a savant, not 'intelligent'. Therefore it can't even be called artificial intelligence.
Laws are rules for the court, but merely a bottom bar to hit for life. Think beyond laws in your actions always.
As you yourself admit, this problem is not limited to AI, so it already exists. And if AI is so accurate that it becomes dangerous to one's job to disagree, then AI can only improve things because the problem *already exists*, but the "preceding diagnosis" will become more accurate with less cost for the subsequent silence.
I would add virtually everything happening in Silicon Valley presently to the list. It's hype built around teenage sci-fi fantasies. That kind of thinking will collide with reality eventually, it's just inevitable. So glad to finally see a backlash (though I suspect that has a little to do with unicorn startups failing to deliver on their fantasies, so once again it comes down to $$$).
The AI isn't self-aware and isn't going to 'know' that it made a mistake, a real human will have to adjust its programming. Sort of like millennials.
Minor quibble - or maybe not minor - but how do you know that the universe is finite? It would be a strange coincidence if the entirety of the universe happened to be the part that we can observe...
W..w..W - Willy Waterloo washes Warren Wiggins who is washing Waldo Woo.
It seems to me "my brain is magic" is more of a religious belief. His "heresies" are all effectively variations on that theme.
Certainly if we were competing for resources, I could understand it somewhat, being more intelligent than us, darwinian evolution predicts that it would survive while we would not, but why should one assume that the resources we need or desire would be the same ones that AI would need or want?
File under 'M' for 'Manic ranting'
All true, and Google/DeepMind never said that AlphaGo, in of itself, is anything other than a dedicated Go playing program. I wouldn't compare it to a calculator though (or DeepBlue for that matter) since it learned how to do what it does (at this point primarily by playing against itself)... nobody programmed it how to evaluate whether a given Go board position is good or bad.
However... DeepMind are in the business of trying to create general intelligence, and are trying to do so based on reinforcement learning coupled with other ML technologies such as neural nets. It's best to think of AlphaGo as a technology demonstrator of the capabilities DeepMind have developed.
Although they havn't done so, I think it would be trivial for DeepMind to create a single general program that could learn to play a range of board games such as Go, Chess, Checkers.
Perhaps from a generalization point of view, another of DeepMind's technology demonstrators is more impressive, where they've got a single program that, via reinforcement learning, has learnt to play dozens of Atari video games, many at super-human level, given nothing more than the screen pixels and current score as input (with zero knowledge of the game or goals other than attempting to maximize the score).
Doesn't anyone else see the death of Moore's Law as the elephant in the room? For the past 50 years, we have seen true exponential growth in processing speed and efficiency, yielding quite a bit of justifiable optimism about AI. But now that we see mounting evidence that Moore's law is either dying or dead, it's only natural to adjust our definition of what's possible. Without exponential growth of processing power, I don't see how we could get to general AI in the next few centuries.
Fast Federal Court and I.T.C. updates
a deep belief that humanity is somehow special in a special way
It's this. Look at pretty much all of SciFi. SciFi is generally very imaginative and rich in outlandish concepts, yet almost without exception humanity not just still exists but it almost always has a very special role to play.
People just have a hard time accepting that humans are not the epitome of what this universe can create. Seeing articles like this is like listening to apes arguing that these weird newfangled homo sapienses will never be dominant because they will never be as strong as orangutans.
RIGHT.
" Specialized artificial intelligence seems to be much easier to achieve and is at the same time likely almost as good as general intelligence for topics such as those. You don't need to develop an artificial general intelligence to cure cancer, if you already developed a specialized artificial intelligence that is able to find a cure."
Here's a further point?
Why the hell would we even want general intelligence in the first place? We want something we can push a button and it solves our problem. Preferably without giving it a lunch break or a vacation or civil rights.
From time to time, I drive a 2016 corvette on Montana highways with 80 mph speed limits. It is fair to say that the car loafs along. It was absolutely built for these speeds, and speeds considerably higher. I often reach those higher speeds. [Um. Allegedly. Cough.] Many other models are built with similar capabilities. The highways here are well designed for those speeds. Even many of the secondary roads here are pretty good for them, though not as good.
Methinks you are thinking well inside your own box. Poorly. Which makes me raise my eyebrows at your assertion that you are a physicist. That may be unfair; many people are notably vertical in their strengths. But still, my eyebrows are raised. :)
We can also (if we are honest) observe that progress, and the potential it unleashes in many cases, is not all that closely linked with what's commercially available or common around the time of the fundamental invention. In the first decade after lasers were invented, for instance, there was no significant commercial application. When the integrated circuit was invented, it wasn't much to look at and functionally speaking, for decades, it was outright pitiful compared to ICs today. We're still dealing with developing a full understanding of how neurons do what they do. In laser parlance, in 2017 we are yet pre-laser, and anyone who tries to tell us that lasers can't do X at this point should be considered, at most, a hand-waver in the grips of a fit of profound hubris.
WRT the subject at hand - intelligence and consciousness resulting from information processing - nature has, fortunately enough, provided numerous models at various levels. So we know it can be done at least one way - neural-like systems. Sure, it's obviously not easy. Brains use very small, very complicated, and very difficult to understand computing elements.
But achieving a manufactured intelligence is also obviously highly interesting and to many, highly desirable. Assuming only that our technological progress doesn't actually halt due to some unrelated factor (war, asteroid, runaway climate, alien invasion, etc.), there are many reasons, all supporting one another very, very well, to assume that we will "get there from here." Not the least of which is there are many (sub-)reasons to presume that will be a great deal of economic leverage in such technology.
And, perhaps most relevant to you, there are no known physics related reasons to presume that we won't get there eventually. As you should know very well. If one is (or multiple are) discovered - for instance, should it be determined at some point in the future that brains use some heretofore unknown physics mechanism(s) to do what they do - then we may quite suddenly be on different grounds in terms of ultimate practicality. But there isn't even a hint of this as yet. It definitely appears to be chemistry, electricity, and topology all the way down as far as brains go. That stuff, we can do. Larger and clumsier and perhaps even slower... perhaps even only as emulation... yet we can do it. We just don't know exactly what to do. Yet.
I've fallen off your lawn, and I can't get up.
Nr 3, that human intelligence can be realized in silicon, or at least in some artificial way is what counts. I would be extremely skeptical of the claim that it is not possible with any future technology that we humans might develop.
Whether artificial AI will be benign or hostile is a big question. But, what's the alternative to AI? Do we want humankind to remain the same for the next million years? Would it be better if we evolved ourselves? Assuming we evolved into a 'homo superior', how many of us alive today would be actual, by blood, ancestors of those superiors and able to claim them as our descendants? If we created AI's 'better than ourselves' would we be able to think of that AI progeny as our children somehow or does there have to be a biological connection?
But really, when people raise objections to superior AI, the main thing I want to them be able to say is what alternatives do they propose?
In theory, theory and practice are the same; in practice they're different. (Yogi Berra & A. Einstein)
You realize the DNC and RNC switched flags several decades ago?
The human population is composed of experts, with divisions of labor. It is not unreasonable for AI programs to have areas of expertise.
In fact this is not true. The human population is composed of experts, some of whom have required in addition specialized skills due to division of labor.
Post may contain irony: discontinue use if experiencing mood swings, nausea or elevated blood pressure.
The situation is not quite similar. Aerodynamics proved that heavier-then-air vehicles would be able to fly due to the effect of lift. So there was theoretical proof, making the actual building of the plane a 'mere' engineering effort.
We also have proved human level intelligence can exist because our species exists. Once again, the situation is nearly identical.
-- All that is necessary for the triumph of evil is that good men do nothing. -- Edmund Burke
The problem with Kevin Kelly is that he tickles the part of your brain that wants more Richard Feynman, and then this.
This thesis is not new.
Kelly on the Future, Productivity, and the Quality of Life — January 2013
From my original notes:
Kelly seems not to comprehend the challenge involved in proving near-equivalency of computational systems (over any ingenious metric) in the finite case. You'd be walking straight uphill in the general direction of Chaitin's constant.
Is lumping omega actually a real problem?
Kelly seems pretty sure that omega comes in flavours marsupial and mammal ("substrates").
Feynman had a supreme knack of not screwing this stuff up, even when he was skirting a field he really didn't know much about. He had such a strong sense of when his own feet were on solid ground, and was extremely clever is turning the discussion to where his solid footing generally carried the day.
Kevin Kelly not so much.
Like my dad said over a decade ago regarding chess computers: "Did people ever stop running because airplanes can go faster? Do runners even care what the speed of an airplane is? Why would chess players care how good machines are at it? Why would it mean anything to philosophy or to competition?"
They're just fancy calculators. Some people get all twisted up whenever the chess computer can make a subtle move, but all it really shows is their lack of understand about the nature of the game. It isn't actually mystical, and humans aren't actually magically better at certain types of calculations or "understanding." Instead, our advantage is that we're generalized and exist in a broad physical context. AI exists in a very very narrow physical context. It can't really compete very well at being human, and humans can't compete very well at being calculators.
As a Big Bang skeptic and Finite Universe skeptic, I wanted to point out that here it doesn't matter; if we agree that most communication will be limited to [the speed of light in a vacuum] and we admit that we're starting from a finite level of intelligence here on Earth, then the growth rate must be limited. Exponential growth of intelligence would only fit within the bounds of the speed of light and the minimum quanta for finite time period and it would eventually have to slow.
Even in an infinite universe, intelligence should remain a somewhat localized phenomenon whose growth is similar to what would be expected in a finite universe, just with more flexible bounds and the possibility of minor outliers.
It comes down to this: the maximum average communication speed is fixed, but as intelligence grows the average distance between nodes increases. Efficiency goes down as the size of the system increases.
The AI isn't self-aware and isn't going to 'know' that it made a mistake, a real human will have to adjust its programming. Sort of like millennials.
If outcomes are an input then it might already be programmed to continually make adjustments, and it might not even need a special mistake-detecting algorithm to achieve it.
For example a traditional PID controller continuously calculates error and attempts to correct for it, and it doesn't use any sort of system of dichotomization to decide if a "mistake" was made. And a PID controller is about the simplest AI there is, you can implement it with gears and springs, or analog electronics, or digital logic. And it is advanced enough to avoid your presumed problem. Presumably, so are millennials.
Another oft told myth.
If you consider a game of Go to be a form of conversation (which is at least arguable), then AlphaGo has already passed the original Turing test.
How do you know we haven't already gotten bored?
I guess that makes it a minor quibble after all :)
W..w..W - Willy Waterloo washes Warren Wiggins who is washing Waldo Woo.
Oh my, look at the buzzwords being trotted out! AlphaGo! Watson! Church-Turning! Driver-less cars! Be still, my beating heart!!
Nay-saying is now the default reasonable approach to the claims of AI true believers. It's not like there isn't some historical context you know. Remember the Giant Brain analogy of the 1950's in reference to Univac? How about how the mainframes of the 1960's, the IBM 360s and 370s? IBM would certainly figure AI out! They are Big Blue!
Next there were the minicomputers of the 1970's and 80's, the PDP, VAX, MAI, Prime, Data General, yeah! Bringing the cost of computing down to the departmental level! So many more AI experts could get their hands on a computer system!
Then it was the personal computers of the 1980's and 90's. Apple, IMSI, IBM, this was the real breakthrough man! Finally hobbyists and garage enthusiasts could attain computing nirvana! Computing for the masses, there's no way AI can be more than 5 years off now!
And in each of these eras, there were AI experts suggesting that "they've finally solved the hard stuff, there's just some detail to fill in now, all we need is a couple of grad students to write the findings up".
Yeah. Empty promises. Way overestimating the AI experts' progress. Way underestimating the hurdles that remained. Impressive bench demos (when compared to zero), but really? This assemblage of string, baling wire and duct tape is supposed to be impressive and take over the world?
There's no complete, accurate, and universally accepted definition of what intelligence even is. Think about that. That's not denial, that's reality. And you're going to solve for a problem that doesn't even have a complete definition? Who's in denial then?
Yes, of course reinforcement learning is intrinsically limited. The current state of the art is deep learning, but even this technique is still statistical machine learning. In other words, it works by devising a cost or objective function (in the case of go, winning the game, which is clearly defined), and optimises it by starting from examples. In the case of the game of go, the hundreds of thousands of games already annotated and played by humans. Reinforcement learning artificially creates many new games from old ones, and playing these, which helps finding a better local minimum in the cost function.
Now "strong" AI or whatever you want to call some version of human-like AI have to cope with situations where there is no clearly defined cost function, and very few if any example to start from. The complexity of these problems is way beyond that of Go, which is a very limited system. Note that these problems are every day ones that humans have to cope with every day, like how do I persuade my clients to pay me ? how do I explain to my work partners the value of my work ? they typically include other complex systems that are not rule-based or for which the rules may exist but are unknown. Example: we do not know enough of physics to accurately predict next week's weather. How do we improve the situation?
You will not be able to better predict next week's weather with reinforcement learning.
"No, it's a narrow AI. In the end, it's simply doing math. It's not "thinking" in any sense of the term."
Oh? How would you define 'thinking' then? Try not to use any weasel words.
The approach is probably more akin to running a network protocol through a fuzzer, looking for exploits. They keep rolling back the savestates until they beat the level recursively, in Super Mario Bros, for example.
I think the mistake that naysayers of AI-apocalypse ignore is that something doesn't have to be superhumanly intelligent, or even humanly intelligent, to be dangerous.
Minor quibble - or maybe not minor - but how do you know that the universe is finite? It would be a strange coincidence if the entirety of the universe happened to be the part that we can observe...
An infinite universe would't be expanding. It couldn't get any bigger.
-- sudon't
Air-ride Equipped
When people say "the universe is expanding", they don't mean in absolute size. It's more like "inflating" - spacetime itself is expanding.
W..w..W - Willy Waterloo washes Warren Wiggins who is washing Waldo Woo.