Slashdot Mirror
When Will AI Surpass Human Intelligence?
destinyland writes "21 AI experts have predicted the date for four artificial intelligence milestones. Seven predict AIs will achieve Nobel prize-winning performance within 20 years, while five predict that will be accompanied by superhuman intelligence. (The other milestones are passing a 3rd grade-level test, and passing a Turing test.) One also predicted that in 30 years, 'virtually all the intellectual work that is done by trained human beings ... can be done by computers for pennies an hour,' adding that AI 'is likely to eliminate almost all of today's decently paying jobs.' The experts also estimated the probability that an AI passing a Turing test would result in an outcome that's bad for humanity ... and four estimated that probability was greater than 60% — regardless of whether the developer was private, military, or even open source."
Never.
I think we heard these exact same words 50 years ago.
Say it ain't so! In other news, Coca-Cola released a statement that in 20 years, more people will be drinking Coca-Cola than there are drinking it now !1!!
I want to delete my account but Slashdot doesn't allow it.
and four estimated that probability was greater than 60%
Of our incredibly small sample size of hand picked Experts, Less than 25% think there is a probably chance! YOU SHOULD BE WORRIED!
Oh come on. I don't even have a computer that can pick up stuff in my room and organize it without prior input, and nobody does, and that would not be close to a general AI when it happens.
They're really assuming that the technology will go from zero to sixty in 20 years. Which they assumed 20 years ago, too, and it didn't happen. Meanwhile, nobody has any significant understanding of what consciousness is. Now, it might be that a true AI computer doesn't need to be conscious, but we still don't know enough about it to fake it. We also have no system that can on demand form its own symbolic system to deal with a rich and arbitrary set of inputs similar to those conveyed by the human senses.
Compare this to things that actually have been achieved: We had the mathematical theory of computation at least 100 years before there was a mechanical or electronic system that would practically execute it (Babbage didn't get his system built). We had the physical theory for space travel that far back, too.
We know very little about how a mind works, except that it keeps turning out to be more complicated than we expected.
So, I'm really very dubious.
Bruce Perens.
One might argue that the fact that the human species wastes so much money (and as a consequence, resources) on fulfilling carnal desires rather than advancing it's civilization, points out that we do not collectively really represent a very high standard of intelligence.
I am certain that another group of 'experts' said the same thing in 1980
Wherever You Go, There You Are
AI research started in the 1950s. Considering how "far" we've come since then, I don't think we should expect any sort of general artificial intelligence within our lifetimes.
People are doing great stuff at "AI" for solving specific types of problems, but whenever I see something someone is touting as a more general intelligence, it turns out to be snake oil.
When the computers are doing all of the intellectual work what will people do? I doubt that factory jobs would be prevalent as the employees would be replaced by robots. Will we simply laze about all day posting on Slashdot? Or, will our robot overlords kill all of us? It seems like the easy solution would be not to develop advanced AI, it's not going to develop itself...yet.
Sent from my iPhone 5
Please define "intelligence."
Calculation speed? An abacus was smarter than humans.
Memory? Not sure who wins that.
Ingenuity? Humans seem to rule on this one. I don't know if I count analyzing every single possible permutation of outcomes as "ingenuity." And I'm not sure we really understand what creativity, ingenuity, etc., really are in our brains.
Consciousness? We can barely define that, let alone define it for a computer.
It seems most people seem to think "calculation speed and memory" when they talk about computer "intelligence."
Should we remember all of the unrealized promises of AI from the 1950's? What makes anyone believe in these baseless claims? If anything, in 20 years they'll give us a better spam filter. Give me a break..
The problem is, this isn't a survey of "AI experts," it's a survey of participants in the Artificial General Intelligence conference. As far as I can see, this is a conference populated by the few remaining holdouts who believe that creating human-like, or human-equivalent, AIs, is a tractable or interesting problem; most AI research now is oriented towards much more specific aspects of intelligence. So this is a poll of a subset of AI researchers who have self-selected along the lines that they think human-equivalent AI is plausible in the near-ish future; it's hardly surprising, then, that the results show that many of them do in fact believe human-equivalent AI is plausible in the near-ish future.
I would be much more interested in a wider poll of AI researchers; I highly doubt anything like as many would predict nobel-prize-winning AIs in 10-20 years, or even ever. TFA itself reports a survey of AI researchers in 2006, in which 41% said they thought human-equivalent AI would never be produces, and another 41% said they thought it would take 50 years to produce such a thing.
No mention, of course, of the new jobs it will make possible. How many web designers were there in 1990? How many airline pilots in 1940?
To me the key word is artificial, depending on your interpretation of the meaning it could be simply man made, or it's fake, simulated.
Does deep blue show any intelligence? To me, that's just good programming. I think the intelligence of computers is a misnomer. Their intelligence so far and has always has been nil. Maybe that'll change, but in so many areas of technology I'm an optimist but in this regard I'm a pessimist or at least very skeptical.
A computer can't even pick a (truly) random number without being hooked up to a device feeding it random noise.
How do you program that? How does the brain choose a random number? What's holding us back? CPU Speed? Quantum computing? A brilliant programmer?
Wake me up when a computer can even do something as simple as pick a truly random number and I'll be impressed.
The most likely scenario is, AI which develops fusion and holographic storage.
When will AI be able to write original jokes that can make people laugh? And how about scripting a funny TV commercial?
And not the wrong kind, either.
Hey don't knock it. If more people wanted some panda-burger, there'd be a lot more of them.
If all else fails, immortality can always be assured by spectacular error.
I'd say "screw it, if it's going to take my job, and jobs of my friends, family and all my descendants, I'm making it a complete dimwit and swearing by all I know that it was impossible to design otherwise, and putting that in every single book and publication on the topic!"
If you have AI smart enough to outsmart people, you probably have something that can learn to control some fairly simple mechanical parts that look like legs and maneuver them based on cheap sensor input and a couple of cameras. So you have robots, who will pretty quickly get the ability to build and maintain themselves. Which means your manual labor jobs go away, too. Which means things like food and raw materials drop to approach the cost of energy. Luckily we'll have some pretty swell solar panels by then, for much cheaper than today, and probably be pretty close to fusion. As energy costs approach zero, the cost of everything in the world approaches zero and requires no human oversight. Everyone will be unemployed and own 40 houses. We can all sit around making YouTube videos of ourselves singing in the hopes that we'll get famous so people will want to have sex with us. It'll be boring, but it won't be the worst thing in the world.
I occasionally attend AI meetings in my local area. The problem with AI development is that too many "experts" don't understand engineering; or programming. Many of today's AI "experts" are really philosophers who hijacked the term AI in their search to better understand human consciousness. Their problem is that, while their AI studies might help them understand the human brain a little better; they are unable to transfer their knowledge about intelligence into computable algorithms.
Frankly, a better understanding of Man's psychology brings us no closer to AI. We need better and more powerful programming techniques in order to have AI; and philosophizing about how the human mind works isn't going to get us there.
No, I will not work for your startup
It matters to the fish who have to share the water with this new beast.
What I mean by that is that I haven't see yet any sign of generic intelligence -- otherwise if you consider programs that beat human at chess "intelligent" that has already happened. But those programs cannot even solve a tic-tac-toe game because they don't actually "understand" what's going on. They have some inputs some processing and they give you an output, if you vary the input and the problem or if you expect a different type of output the program would not know how to adjust, therefore I would not considered that "intelligent". Neuronal nets and artificial brains are another thing, but they are still at the very beginning.
"superhuman intelligence" there might be some limit to intelligence, I don't mean memory and computation speed, I mean the understanding that if "A implies B" then "non B implies non A"... once an artificial brain understands that concept there's not so much more to understand about it.
"It is our choices, Harry, that show what we truly are, far more than our abilities." -- Prof. Dumbledore
I think the OP found some news piece from the 1960s and decided to recycle it...
If I was to make such a prediction I would say 60 years, then even if it would not happen I would not be around to be shamed.
AI working? Nooooo! (and I'm in a machine learning research group)
metageek
Awesome quote. The stuff people imagine is hysterical. For a robot to evolve a free will, it will be given to him by humans, so in essence it is not free at all. If robots are evil, it is because people are inherently evil and program it to think methodically instead of compassionately. It is easy to program the functionalism of a human mind, but behaviorism will never be fully understood. Computers are already superhuman in many ways, but to compose music, write classic literature, cook lavish meals, it will never ever happen. Keep dreaming dreamers. Any creativity a robot contains would have come from our own instruction.
If it doesn't have emotions and we mistreat it then it will logically see us as a threat to its own survival and try to eliminate us.
I agree with many of your sentiments, but I think they're still too anthropocentric. We evolved in an environment where survival was very nearly the prime directive (just after "pass along your genes"). Strong AI will be developed in a lab. We could create the "smartest" computer in the world, but who would feed it goals, and the lengths it would go to achieve those goals?
If an AI is tasked with finding a Theory of Everything, and someone decides to take an axe to its circuits, will it determine that the axe is a threat to its goal, and act accordingly? Or will it simply interpret it as another in a long series of alterations to its circuits? Or perhaps it will ignore it altogether, considering it irrelevant.
Because ultimately, those options were programmed in by a human. Our strong AI - the first ones at least - aren't going to be independent life forms with their own dreams and desires. They will be tools to help us solve problems, and I think they will be well-understood by many, many computer scientists. When something unexpected happens, the program will be debugged, and altered to prevent the unexpected behaviour.
If there is a robot apocalypse, it won't be because we didn't treat our creations right, but because some 13-year-old hacker in Russia said "I wonder what happens if I do this".
Last post!
Do you humor chimpanzees for leading to our evolutionary branch? Sure we keep a few in zoos, but what about all the others whose habitat we regularly destroy for our own selfish needs? Do you think an AI would act differently? Is there some threshold of intelligence where suddenly you care about humoring ants, or does nearly everyone not pay them any attention and try to eradicate them if they get in the way or become pests?
The point was that subs move fast than fish but they don't swim. Similarly computer do tons and tons of things faster than people but don't think. (That's how I read it anyways)
When it comes to predicting the impact of a sentient AI on human civilization, there is never any shortage for alarmism. I am not an expert, but I am a programmer. And I believe three things to be true with respect to AI.
1) Until we have a better understanding of why humans are sentient in the first place, we are probably not going to get any closer to recreating that phenomenon in a computer program.
2) A Turing Complete AI is about as far off as the discovery of a room temperature super conductor or a form of fusion suitable for large scale power generation. We may be close, but probably not *that* close.
3) I seriously doubt that any AI that we are going to be able to create with anything resembling current computer technology is going to have a thought process even close to our own.
Think about it for a moment. Human intelligence is shaped as much by our 5 senses, our capability to create and understand language, our emotions, our ability to affect our surroundings and observe those effects, and to communicate with one another as it is our capability for logic and math. The factors that will shape an A.I. are so different as to create the possibility that a Human Intelligence and an Artificial Intelligence may not even be able to meaningfully communicate.
Will the first sentient AI be hosted on a single computer, or will it be a gestalt effect encompasing the entire internet?
Will the sentient AI be aware of time in anything even close to the way that we are?
Will the sentient AI even be capable of 'wanting' anything, given that it will have no need for sleep?
Will the sentient AI be able to comprehend the nature of its existence as a program, and be able to manipulate its own variables by choice?
Will the sentient AI fear its own termination, or not really care knowing it can easily be reloaded?
I would say that being threatened by a computer based AI that is better able to perform 'intellectual work' is about as reasonable as being threatened by cheetah's because they are better at running really goddamn fast.
I will admit that the idea of AI's eliminating paying jobs of a particular sort is an interesting problem to consider, but not that different from considering what will happen when we can create robots capable of performing all types of manual labour. Will that result in world wide poverty, or will it result in world wide prosperity ala StarTrek?
END COMMUNICATION
Which three men in a tub assumed 20 years ago, too, and it didn't happen.
The first rule of thumb is never to believe a prediction by anyone who writes grant applications for a livelihood, which covers most living scientists.
Computers will acquire a patchwork of amazing abilities over the next three decades. I'm not sure it's particularly useful to measure this against a three year old. Right now we're further along on "fly airplane" than "tie shoes". If there was a Turing test to declare whether a task is simple or not, humans would fail.
A Google data center with 100,000 CPU nodes is already pretty far up the cognitive scale, but it's not a form of cognition we've bothered to define as such. The most important intelligence will be assisted intelligence: what humans accomplish in collaboration with their tools. The tools will become increasingly amazing, at first on a patchwork basis, and then the seams will become increasingly unclear.
Right now social networking sites predict what we might find interesting on fairly trivial low-dimensional criteria. Netflix must be the all-time champion of the drunken I-fought-with-my-wife-tonight 1-5 rating. Could the data set possibly be less rich or more corrupt? And already we squeeze something out. Just wait until the computers know everything about us and the ability of the computer/network to anticipate our cognitive whims becomes spooky prescient.
On another front, some of the fruits of neurology are now coming on line. I have no idea whether this stuff works or not. Typical how we trip over our own shoelaces, trying to get speech recognition to work *before* mastering auditory grouping, which strikes me as far more fundamental.
From Audience based on research by Lloyd Watts
Audience is the first company to deliver a commercial product based on the science of [a]uditory [s]cene [a]nalysis, which entails the grouping of components in a complex mixture of sound into sources. Just as the human auditory system can readily ignore background noises while focusing on a voice of interest, [our stuff achieves] noise suppression up to 30 dB for both stationary and non-stationary noise sources to provide [adjective of awesomeness] voice quality within even the [pertinent superlative].
Just going to point out we already have holographic storage. There is just no commercial products that do it yet. Contrast with fusion and Turing test passing AI.
We already have fusion. It just costs more energy than it generates. It's just a matter of increasing the efficiency a bit more. Contrast that with Turing test passing.
What else do you expect a computer to do if not execute an algorithm? It's a *computer*. It *computes*. The question is if a large enough collection of such simple algorithms interacting with each other will create the illusion of intelligence. When it comes to chess the answer is yes, so much so that Kasparov insisted Deep Blue was fed moves by a human.
You speak of a mythical "intelligence" when for all we know we ourselves might well be a collection of simple algorithms being executed in our heads.
When will AI surpass human intelligence? As soon as we figure out how to do artificial intelligence the way popular culture conceives of it.
There are two main areas of AI research, as I see it:
(1) Engineered intelligence. These systems learn, but they learn in carefully controlled structures, like Markov models and mapping functions in genetic algorithms.
(2) Emergent intelligence. These are based on evolving systems of simpler structures, like neural nets, and those little cooperating robots you keep hearing about. In some ways, since the intelligent behavior evolved over time, this is more akin to natural intelligence then artificial intelligence.
Neither group has really accomplished a hell of a lot. Speech recognition and computer vision still suck ass. Group (1) has been dominant since the idea of AI was developed, and frankly, they're not a millimeter closer to understanding how to build up a system that is intelligent, where you understand all the parts you built with. Group (2) is making some progress, but then they're left with a system they don't understand because they didn't engineer it.
Dorks like Kurtzweil seem to think that as soon as we can fit as much compute power into one chip as we GUESS is in the brain, we'll magically get sentient robots. That's bullshit. We need software systems that learn and adapt, and we just haven't figured out how to make those.
You're right about his romanticizing what life was like 100 years ago. I need to kick back and watch TV and have a cold soda from the fridge after work. I also want to take a hot shower when I get home. On the weekends I might enjoy camping or fishing. None of those were available 100 years ago. Life was pretty bleak unless you were one of the robber barons. But 40 years ago, Mom was at home. Dad put in a 40 hour week at the factory. The working class was entitled to a pretty good share of the wealth that they were creating. Now between Mom and Dad, the family puts in 80+ hours on the job. College degrees just to have comparable living standards. Where the hell is my flying car? Where did we go wrong?
At least part of today's 10% unemployment rate stems from the fact that we use machines to do what people used to do. Imagine how many of us will be unemployed when we don't need any human beings who can think. How will you earn a living then?