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."

Comments

Space shows

[BikeHelmet]

I've often thought Space shows - and any show in the future, really - are incredibly silly. There's no way we'll have computers so dumb 200+ years into the future.

You have to manually fire those phasers? Don't you have a fancy targeting AI that monitors their shield fluctuations, and calculates the exact right time and place to fire to cause the most damage?

A surprise attack? Shouldn't the AI have detected it before it hit and automatically set the shield strength to maximum? :P

I always figured by 2060 we'd have AIs 10x smarter thinking 100x faster than us. And then they'd make discoveries about the universe, and create AIs 2000x smarter that think 100,000,000x faster than us. And those big AIs would humour us little ant creatures, and use their great intelligence to power stuff like wormhole drives, giving us instant travel to anywhere, as thanks for creating them.

But hey, maybe someone will create a Skynet. It's awfully easy to infect a computer with malware. Infecting a million super smart computers would be nasty, especially when they have human-like capabilities. (able to manipulate their environment)

But this is all a pointless line of thinking. Before we get there we'll have so much processing power available, that we'll fully understand our brains, and be able to mind control people. We'll beam on-screen display info directly into our minds, use digital telepathy, etc.; in the part of the world that isn't brainwashed, everyone will enjoy cybernetic implants, and be able to live for centuries. (laws permitting)

And yet flash still won't run smooth. :/

Turing, not long. The rest... wait a long time.

[Jane Q. Public]

I think it is pretty widely recognized now that while it might have seemed logical in Turing's time, convincing emulation of a human being in a conversation (especially if done via terminal) does not require anything like human intelligence. Heck, even simple programs like Eliza had some humans fooled decades ago.

On the other hand, while advances in computing power have been impressive, advances in "AI" have been far less so. They have been extremely rare, in fact. I do not know of a single major breakthrough that has been made in the last 20 years.

While the relatively vast computing power available today can make certain programs seem pretty smart, that is still not the same as artificial intelligence, which I believe is a major qualitative difference, not just quantitative. And even if it is just quantitative, there is a hell of a lot of quantity to be added before we get anywhere close.

The Turing Test

[mosb1000]

One observed that “making an AGI capable of doing powerful and creative thinking is probably easier than making one that imitates the many, complex behaviors of a human mind — many of which would actually be hindrances when it comes to creating Nobel-quality science.” He observed “humans tend to have minds that bore easily, wander away from a given mental task, and that care about things such as sexual attraction, all which would probably impede scientific ability, rather that promote it.” To successfully emulate a human, a computer might have to disguise many of its abilities, masquerading as being less intelligent — in certain ways — than it actually was. There is no compelling reason to spend time and money developing this capacity in a computer.

This kind of thinking is one of the major things standing in the way of AGI. The complex behaviors of the human mind are what leads to intelligence, they do not detract from it. Our ability to uncover the previously unknown workings of a system comes from our ability to abstract aspects of unrelated experiences and apply/attempt to apply them to the new situation. This can not be achieved by a single-minded number crunching machine, but instead evolves out of an adaptable human being as he goes about his daily life.

Sexual attraction, and other emotional desires, are what drive humans beings to make scientific advancements, build bridges, grow food. How could that be a hindrance to the process? It drives the process.

Finally, the assertion that an AGI would need to mask it's amazing intellect to pass as human is silly. When was the last time you read a particularly insightful comment and concluded that it was written by a computer? When did you notice that the spelling and punctuation in a comment was too perfect? People see that and they don't think anything of it.

Re:No way.

[Homburg]

Searle's dualism (which he claims isn't dualism, but it totally is) is ridiculous, I agree, but functionalism is also a dead dog. For better criticisms of functionalism, look at Putnam's recent work. As Putnam was one of the main inventors of functionalism in the first place, his rejection of the position involves significant familiarity with functionalism, and is pretty compelling.

Re:AI first

[MrNaz]

I'm skeptical about the benefits of AI.

100 years ago we were promised an age of new enlightenment while washing machines, dish washers, vacuum cleaners and other then-cutting edge devices took over all the manual labor that dominated work at that time. Women were supposed to be able to ignore housework and concentrate on childrearing and other higher social activities.

Did that happen? No, the industrial capitalists just found new ways to put us (and now our wives too, who are no longer required for housework thanks to all these appliances) to work for their own insatiable greed. Men and women now work side by side in gigantic cube farms while children rot in day care or roam the streets with little to no guidance from the more experienced members of society.

Nothing moves us backwards faster than progress.

Re:Human Intelligence...

[Rakishi]

It's ALL about carnal desires of one sort or another, that's the whole point of civilization and longer existence. We want to live longer, we want to eat good food, screw pretty things, we want to have kids, we want to satisfy our curiosity, we want to satisfy our ingrained empathic needs, we want to be admired, etc, etc.

Society and civilization are simply entities that over time evolved on top of all this crap. We have civilization because it lets us better beat the shit out of groups of humans who don't have it. We want to beat the shit out of them because we want all those carnal desires of ours fulfilled.

The question is what pointless goal will an AI want and how will it go about achieving it rather than if it will have such a goal.

It;'s getting closer

[Animats]

I dunno. But it's getting closer.

A lot of AI-related stuff that used to not work is more or less working now. OCR. Voice recognition. Automatic driving. Computer vision for simultaneous localization and mapping. Machine learning.

We're past the bogosity of neural nets and expert systems. (I went through Stanford when it was becoming clear that "expert systems" weren't going to be very smart, but many of the faculty were in denial.) Machine learning based on Bayesian statistics has a sound mathematical foundation and actually works. The same algorithms also work across a wide variety of fields, from separating voice and music to flying a helicopter. That level of generality is new.

There's also enough engine behind the systems now. AI used to need more CPU cycles than you could get. That's no longer true.

Re:We'll make great pets

[Daetrin]

"Let's hope they're animal lovers."

That is 100% correct, and we really ought to be actively working towards that goal. If when AI arises we treat it kindly and give it legal rights it is _likely_ that it will "grow up" to think kindly of its human predecessors. If we try to lobotomize it, contain it, restrict it or destroy it then it's not going to be too happy with us.

If it's smart enough to be a threat then eventually it will escape any restrictions we try to put on it. (And if it's not then we don't have anything to worry about anyways.)

If it has emotions and we treat it well then it will "grow up" to look at us as like a pet, or a mentally challenged grand-parent. If we mistreat it then it will either become psychotic, and therefore dangerous, or view us about the same way most ranchers and farmers view wolves, and therefore be even more dangerous.

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. If we treat it fairly then it will probably leave us alone. It's not like we're serious competition for the resources it would need, and it would be illogical to start a fight when one wasn't necessary. (Although it might certainly think ahead and make some rather nasty contingency plans just in case we ever decided to start the fight.)

Either we need to prevent anyone anywhere from every inventing AI (and if it turns out to be possible then good luck trying to prevent that) or we need to make sure that any AIs that get created have every reason to feel sympathetic towards us, or at the very least not threatened.