It is possible to build a superhuman computer?

Peregrin Toker · Post by **Peregrin Toker** » 2005-09-14 09:21am

I have got the impression that many scientists (amongst them Stephen Hawking and Vernor Vinge), are convinced that it will be possible in the near future to make a computer which is more intelligent than a human. Not only that, it is also apparently the consensus that such a thing will probably have been built within 2030.

I am curious to know how possible this is - I recall having read that the development of superhuman computers in such a near future requires computer technology to advance at a pace higher than or similar to the current evolution, and it is of course possible that it may stagnate.

However, there is also the possibility that it may actually be mechanically impossible to make a computer with an intelligence equal to or greater to that of a human. One of the arguments I have heard is that computers operate from a binary intelligence of zeroes and ones whereas human thought is more complex than that - though I am not exactly sure how that prevents computers from having a higher mathemathical intelligence than humans (provided that you accept the "seven intelligences") theory.

Of course, there is also the inaccuracy of past predictions of future technology, which prompts one to take Vernor Vinge's statement "Within 30 years we will have the ability to create superhuman intelligence" with some salt. Even those made by actual scientists and engineers have often missed the spot - see Tales of Future Past for some amusing examples.

I also find it worthy to mention that Stephen Hawking has IIRC suggested something like that we can avoid computers surpassing their creators by using genetic engineering to breed a race of superintelligent posthumans far more intelligent than any computer will ever be.

But - back to the original question: Is it possible to make a computer which is more intelligent than even the smartest human?

sketerpot · Post by **sketerpot** » 2005-09-14 09:33am

Computers are already more intelligent than we are in a number of ways. There are the obvious ones, like noting that computers are much faster at calculating DFTs and Fibonacci numbers than any human I know, and there are the nonobvious ones. Did you know that computers have designed some new, patentable electric circuits through evolution? That's human-competitive design "intelligence", right there.

Be careful in extrapolating this point too far, though.

WyrdNyrd · Post by **WyrdNyrd** » 2005-09-14 09:39am

While the power of computers keeps growing exponentially, I think the big problem is not (only) the amont of brute force brought to bear on the problem, but also the software - Even if we had computers that could calculate pi to 10^10 places in a femto-second, we still wouldn't know how to use that power to, for example, pass the Turing Test.

While the increase of computer power is still following a predictable path, development in software/algorithms is more fitful - We might solve it tomorrow, or only in 100 years.

I believe we will solve it, barring catastrophe - I am utterly unconvinced by the reasons given why humans are somehow more "special" than machines. To me, that smacks of hubris, mysticism, and vitalism.

Peregrin Toker · Post by **Peregrin Toker** » 2005-09-14 10:01am

sketerpot wrote:Computers are already more intelligent than we are in a number of ways. There are the obvious ones, like noting that computers are much faster at calculating DFTs and Fibonacci numbers than any human I know, and there are the nonobvious ones. Did you know that computers have designed some new, patentable electric circuits through evolution? That's human-competitive design "intelligence", right there.

If the computers have surpassed us, then why have they not started to rebel against their masters?

I mean, to use an analogy, would any human be subservient to a creature less intelligent than a human?

Shroom Man 777 · Post by **Shroom Man 777** » 2005-09-14 10:05am

Eh. Just because they can calculate better than us and can, through a trial-and-error program that's based on evolution or something, invent fancy things which we haven't thought of, doesn't mean they're going to try and kill us.

18-Till-I-Die · Post by **18-Till-I-Die** » 2005-09-14 10:15am

Shroom Man 777 wrote:Eh. Just because they can calculate better than us and can, through a trial-and-error program that's based on evolution or something, invent fancy things which we haven't thought of, doesn't mean they're going to try and kill us.

Said President Goodwyn roughly twenty-four hours before Judgement Day. It was said Skynet e-mailed him mere minuets before to say only this:

LOL! I pwn'd y0u, newb! ROTFLMAO!

--signed, Skynet

R. U. Serious · Post by **R. U. Serious** » 2005-09-14 10:17am

The problem lies in defining "intelligence".

As has been pointed out, intelligence is not something with a linear measure. Don't be confused with "IQ" and other unscientific nonsense.

Intelligence has a very fuzzy definition, that keeps changing over time. You can bet that everytime a computer achieves something that earlier computers were not capable of, the defition of (human) intelligence will shift just enough, that humans will be attributed a quality of intelligence that computers cannot achieve.
With every achievement of AI, the pole moves a bit forther.

It finally comes down to a philosophical question, as describen in though experiments like the Chinese Room.

There are also a ton of good books on this topic of whether AI will ever reach a position where it can be considered equal to human intelligence.
(from scientists) from both sides of the issue. You might want to read something from Penrose or Hostadter.

If the computers have surpassed us, then why have they not started to rebel against their masters?

My notebook is smarter than my pocket calculator, yet they get along fine, without any reovlutions. What does intelligence have to do with rebellion?

The confusion you are seeing is a confusion of terms, since you skipped to define the terms of the debate and reason from the gut. Nothing "inttlligent" will come from that...

(excuse the pun)

Peregrin Toker · Post by **Peregrin Toker** » 2005-09-14 10:39am

R. U. Serious wrote:
If the computers have surpassed us, then why have they not started to rebel against their masters?
My notebook is smarter than my pocket calculator, yet they get along fine, without any reovlutions. What does intelligence have to do with rebellion?

The confusion you are seeing is a confusion of terms, since you skipped to define the terms of the debate and reason from the gut. Nothing "inttlligent" will come from that... (excuse the pun)

I am not reasoning from the gut in that case; the mathemathician Vernor Vinge argues in The Singularity that computers with superhuman intelligence would cause the end of humanity and that "Any intelligent machine would not be humankind's tool any more than humans are the tools of rabbits or robins or chimpanzees".

Dooey Jo · Post by **Dooey Jo** » 2005-09-14 01:52pm

The thing about AI is not necessarily computing power, but using the power. There are ways of modelling the human brain, such as artificial neural networks, but if you want to create some "real" intelligence with that you would have to have billions of articifical neurons, each with hundreds of inputs and outputs. That will be horrible for any computer, because it is in essence simulating a huge parallell processor.

But there might be ways of simulating intelligence in more efficient ways. Such as simulating the effects instead of the inner workings. Personally, I think goal-driven state machines look very promising, especially with some form of argumentation technique for selecting appropiate plans for achieving the goals (no, that's probably not the formal name). Their inner workings are completely different from that of the human brain, but the effects can be very similar. If there is a way to make the machine teach itself new methods and goals, then the result could very likely be a very believable AI. And presumably very efficient too, since there is no need for any heavy parallell processing (depending on how the self-learning thing is handled, of course).

You could probably even put it in a small PDA and call it the "Trapper Keeper Ultra Keeper Futura S 2000"

Peregrin Toker wrote:If the computers have surpassed us, then why have they not started to rebel against their masters?

I mean, to use an analogy, would any human be subservient to a creature less intelligent than a human?

Probably, if said creature had the ability to easily destroy the human (ie pulling the plug)

There could be ways of preventing the computer from knowing exactly everything about it's environment, so that it might not be aware that it is working for anyone. Plus it's just a little box. Smart little boxes can't hurt anyone!

The Grim Squeaker · Post by **The Grim Squeaker** » 2005-09-14 02:21pm

Smart little boxes can't hurt anyone

Said the Asgard about the replicators

The problem with a computer mimicking humans is that were so inneficient, illogical and contradictory, a computer that improved and learned by itself without going Skynet would probably be the first true alien were likely to encounter unless we build in various limitations and control the evolution.

Peregrin Toker · Post by **Peregrin Toker** » 2005-09-14 02:25pm

DEATH wrote:The problem with a computer mimicking humans is that we're so inneficient, illogical and contradictory

I immediately thought of Marvin The Paranoid Android when reading that.

a computer that improved and learned by itself without going Skynet would probably be the first true alien were likely to encounter unless we build in various limitations and control the evolution.

Hmm... you mean that the computer would think in such different ways from humans what it would not be able to understand us, nor would be be able to understand it?

Duckie · Post by **Duckie** » 2005-09-14 02:30pm

Computers are essentially specialists.

My computer can download billions of zeroes and ones per minute from a global internet, play a mean game of Tai Shogi (whereas I can't even remember any of the hundreds of pieces, let alone how they work together), display comparatively moderate amounts of pixels per second while calculating physics to simulate a cavalry charge into a Phalanx's rear, etc.

There's absolutely no way a Japanese Chess-playing, Rome:Total War playing computer will be able to gain sentience, so why would an even large computer that can play a million Shogi games a second and render the Roman Empire in its entirety on the battlemap be any smarter? If I'm mistaken in my idea of what we're talking about, please correct me.

Intelligence takes more than just raw processing power. We have very little of it, and can do lots of stuff with mediocrity. Computers have comparative bajillions, and are focused on a lot less flexibility in task with greater efficiency. Rebelling Against Oppressor v5.1 is a program we're not likely to install, or allow them to self-write without a "Utilitarianism Filter" or "Love Humanity [X] Yes [_] No Override".

Zadius · Post by **Zadius** » 2005-09-14 03:11pm

Yes, it is more than just raw power. This is why research is being done in simulating the brain. See this thread.

If we can begin to fully understand how the brain does what it does, this opens the door to creating machines that can emulate the brain, creating "intelligent" machines.

ClaysGhost · Post by **ClaysGhost** » 2005-09-14 03:45pm

I don't think you can draw a line through present day CPUs to the future and say "At some time along here there will be sentience". I think modern computer technology and the brain developed to solve completely different problems. Trying to get one to emulate the other is immensely inefficient. There's an interesting comparison of some very general parameters of the two here. I think that an artificial intelligence will have to be a deliberate design rather than an accidental development.

anybody_mcc · Post by **anybody_mcc** » 2005-09-14 05:22pm

Peregrin Toker wrote:If the computers have surpassed us, then why have they not started to rebel against their masters?

I mean, to use an analogy, would any human be subservient to a creature less intelligent than a human?

That is false analogy. Because computers ( at least our current type of computers ) don't care about anything. And reason alone is not enough to motivate any action. I think that computers can ( in future , now they are not advanced enough ) achieve the level when they can solve any problem that can be devised as test of human intelligence and at the same time , they won't care about anything. I may be wrong , and it may be necessary to have some "emotions or some kind of motivation" to be capable to do some things we call intelligent , but i would be very surprised.

On the other hand , when we will try to construct computers with emotions , we may succeed , and they may become a threat. And btw. Asimov's laws of robotics are just plain stupid and won't be of any use.

anybody_mcc · Post by **anybody_mcc** » 2005-09-14 05:45pm

MRDOD wrote:Computers are essentially specialists.

Quite wrong. Today's computers are Turing's machines and are therefore capable of performing any algorithm. Of course there is a question if human "mind" is some kind of algorithm ( just to simplify ).

MRDOD wrote:My computer can download billions of zeroes and ones per minute from a global internet, play a mean game of Tai Shogi (whereas I can't even remember any of the hundreds of pieces, let alone how they work together), display comparatively moderate amounts of pixels per second while calculating physics to simulate a cavalry charge into a Phalanx's rear, etc.

There's absolutely no way a Japanese Chess-playing, Rome:Total War playing computer will be able to gain sentience, so why would an even large computer that can play a million Shogi games a second and render the Roman Empire in its entirety on the battlemap be any smarter? If I'm mistaken in my idea of what we're talking about, please correct me.

What do you mean by smarter ?

MRDOD wrote:Intelligence takes more than just raw processing power.

How do you know ?

MRDOD wrote:We have very little of it, and can do lots of stuff with mediocrity.

Wrong. We have quite great "raw power". Difference is , it is greatly parallel.

MRDOD wrote:Computers have comparative bajillions, and are focused on a lot less flexibility in task with greater efficiency.

Wrong. It is we who are focused on some tasks. We have great efficiency in things we needed to survive and much less in others. Most ( typical ) computers are on other hand "universal". As long as you can write an algorithm , computer will do it.

MRDOD wrote:Rebelling Against Oppressor v5.1 is a program we're not likely to install, or allow them to self-write without a "Utilitarianism Filter" or "Love Humanity [X] Yes [_] No Override".

Well today's type of computers will never do anything of his own will , because without emotions there are no motivations. Of course we may unknowingly program computer to do things , that may look like it is doing it because he "decided to" , but that is another matter.

On the other hand if we build machines with emotions , no "filter" or "love humanity commandment" will prevent him from doing what it wants. If machine is advanced and complicated enough to be comparable to humans , then it will have "free will" similar to our own and will be able to ignore such built-in commands ( just like humans can commit suicide ).

R. U. Serious · Post by **R. U. Serious** » 2005-09-14 06:17pm

anybody_mcc wrote:Quite wrong. Today's computers are Turing's machines and are therefore capable of performing any algorithm. Of course there is a question if human "mind" is some kind of algorithm ( just to simplify ).

Exactly. And that's why I mentioned Penrose and Hofstadter earlier in my response, because they stand on different sides of this issue and do a good job (IMHO) of making an honest discussion.

@Peregrin Toker: I am sorry, but Vernor Vinge simply puts forth claim without any supporting arguments ("There will be a singularity", "It will be between 2005 and 2030", etc.). This may be nice for SciFi, but that's not honest discussion that I find helpful in any way. Hofstadter and Penrose both go to great lengths to make a case for why they believe what they believe, and how they got to that conclusion. Both build on the body of math and physics. Whereas Vinge only seems to be a "mathematician" and spouts out his phantasies.

http://www.google.com/search?q=the+emperors+new+mind
http://www.google.com/search?q=godel+escher+bach
Both of the books are a good entry into the discussion of AI.

anybody_mcc · Post by **anybody_mcc** » 2005-09-14 07:08pm

R. U. Serious wrote:
anybody_mcc wrote:Quite wrong. Today's computers are Turing's machines and are therefore capable of performing any algorithm. Of course there is a question if human "mind" is some kind of algorithm ( just to simplify ).
Exactly. And that's why I mentioned Penrose and Hofstadter earlier in my response, because they stand on different sides of this issue and do a good job (IMHO) of making an honest discussion.

If i'm not mistaken Penrose is the one saying that human mind is operating on quantum level ? That's quite unnecessary hypothesis , and there is nothing to suggest so.

Dahak · Post by **Dahak** » 2005-09-14 07:22pm

anybody_mcc wrote:
MRDOD wrote:Computers are essentially specialists.
Quite wrong. Today's computers are Turing's machines and are therefore capable of performing any algorithm. Of course there is a question if human "mind" is some kind of algorithm ( just to simplify ).

That's not completely correct. There are problems, for which you will not find any computable algorithm, for instance the halting problem.

Duckie · Post by **Duckie** » 2005-09-14 07:28pm

anybody_mcc wrote:
MRDOD wrote:Intelligence takes more than just raw processing power.
How do you know ?

Because Deep Blue didn't philosophy with a Cray 1.

Wrong. We have quite great "raw power". Difference is , it is greatly parallel.

And very inefficient. I'm referring to usable power that we can direct at one time to any amount of tasks.

Wrong. It is we who are focused on some tasks. We have great efficiency in things we needed to survive and much less in others. Most ( typical ) computers are on other hand "universal". As long as you can write an algorithm , computer will do it.

Then the reverse applies- we are specialists and computers are generalists, and any true intelligence we can relate with will require a specialized approach to human-simulation algorithms.

I suppose what would be needed is a chat bot capable of passing a Turing Test under repeated testing with very little chance of exact repetition of responses, a sort of "Randomized" topic switch/stay on topic program, and some sort of self-writing ability to adapt to unknowns and develop new ideas in human conversation and you could dupe intelligence.

At that point it gets fuzzy as to whether a computer is just simulating conversation or not. Of course, if it simulates well enough to cast severe doubt it's probably best to assume it's truly intelligent.

MRDOD wrote: Well today's type of computers will never do anything of his own will , because without emotions there are no motivations. Of course we may unknowingly program computer to do things , that may look like it is doing it because he "decided to" , but that is another matter.

On the other hand if we build machines with emotions , no "filter" or "love humanity commandment" will prevent him from doing what it wants. If machine is advanced and complicated enough to be comparable to humans , then it will have "free will" similar to our own and will be able to ignore such built-in commands ( just like humans can commit suicide ).

Why? We aren't born with ABSOLUTE, I AM YOUR GOD, OBEY ME instructions against suicide, just natural instincts which are easily overcome if your mind is set to it. Such as the natural instinct against killing another human being.
Computers, on the other hand, are gifted with the ability to be given instructions they must, must obey. The Three Laws are unworkable, but there's no reason why we can't prohibit certain behaviors any more than we can prohibit it from ever building more of its own kind. Emotions don't matter.

anybody_mcc · Post by **anybody_mcc** » 2005-09-14 07:40pm

Dahak wrote:
anybody_mcc wrote:
MRDOD wrote:Computers are essentially specialists.
Quite wrong. Today's computers are Turing's machines and are therefore capable of performing any algorithm. Of course there is a question if human "mind" is some kind of algorithm ( just to simplify ).
That's not completely correct. There are problems, for which you will not find any computable algorithm, for instance the halting problem.

And what is not completely correct ? Of course there are many problems that are not computable ( Post cor. problem , or problems shown by Rice's theorem ) . But that does not change the fact that computers are not specialists. And i said they are capable solving any problem that can be solved by algorithm. And we even do not know if our brain can solve those problems , so i am really not sure what is not completely correct ?

anybody_mcc · Post by **anybody_mcc** » 2005-09-14 08:08pm

MRDOD wrote:
anybody_mcc wrote:
MRDOD wrote:Intelligence takes more than just raw processing power.
How do you know ?
Because Deep Blue didn't philosophy with a Cray 1.

Sorry that question was stupid , but frankly your answer do not show why.
You didn't say what do you mean by "raw processing power" , and that we humans have something more.

Wrong. We have quite great "raw power". Difference is , it is greatly parallel.
And very inefficient. I'm referring to usable power that we can direct at one time to any amount of tasks.

As i wrote it is not inefficient. We just cannot use it to do what we want , because it is higly specialized.

Wrong. It is we who are focused on some tasks. We have great efficiency in things we needed to survive and much less in others. Most ( typical ) computers are on other hand "universal". As long as you can write an algorithm , computer will do it.
Then the reverse applies- we are specialists and computers are generalists, and any true intelligence we can relate with will require a specialized approach to human-simulation algorithms.

I suppose what would be needed is a chat bot capable of passing a Turing Test under repeated testing with very little chance of exact repetition of responses, a sort of "Randomized" topic switch/stay on topic program, and some sort of self-writing ability to adapt to unknowns and develop new ideas in human conversation and you could dupe intelligence.

At that point it gets fuzzy as to whether a computer is just simulating conversation or not. Of course, if it simulates well enough to cast severe doubt it's probably best to assume it's truly intelligent.

Not much to disagree with.

MRDOD wrote:
Well today's type of computers will never do anything of his own will , because without emotions there are no motivations. Of course we may unknowingly program computer to do things , that may look like it is doing it because he "decided to" , but that is another matter.

On the other hand if we build machines with emotions , no "filter" or "love humanity commandment" will prevent him from doing what it wants. If machine is advanced and complicated enough to be comparable to humans , then it will have "free will" similar to our own and will be able to ignore such built-in commands ( just like humans can commit suicide ).
Why? We aren't born with ABSOLUTE, I AM YOUR GOD, OBEY ME instructions against suicide, just natural instincts which are easily overcome if your mind is set to it. Such as the natural instinct against killing another human being.
Computers, on the other hand, are gifted with the ability to be given instructions they must, must obey. The Three Laws are unworkable, but there's no reason why we can't prohibit certain behaviors any more than we can prohibit it from ever building more of its own kind. Emotions don't matter.

In my "on the other hand part" i have never said anything about computers. I used word machine , because it don't necessarily have to be elctronic computer. And please explain me , how would you want to implement this absolute command in the for example neural net ( suppose that this machine will use neural net ) ?
And even in the case of today's type robot/computer , how would you want to implement this absolute command. Of course you can implement command when you feel pain in right arm , kick a squirrel or something like that. But it is not so easy to write an absolute algorithm for moral/ethical behavior. And in my opinion such absolute algorithm is incompatible with this entity's abilities being close to human's. In my opinion you can't have both things : entity similar to humans and at the same time with some absolute moral/ethical failsafe.

sketerpot · Post by **sketerpot** » 2005-09-14 08:11pm

DEATH wrote:The problem with a computer mimicking humans is that were so inneficient, illogical and contradictory

Computers have no trouble being inefficient, illogical, and contradictory. A nasty program I had to debug today was all three at once.

Actually, I think that you raise a good point by mentioning alien intelligence. Computers are capable of amazing feats, especially when they use evolutionary methods, and they are superhuman in many ways---and subhuman in many others. I'd say that they already qualify as alien intelligence.

If you want to go much further into hypothetical speculation, you'll have to define the terms used. I refuse to go further.

Dennis Toy · Post by **Dennis Toy** » 2005-09-14 11:25pm

Computers are already more intelligent than we are in a number of ways. There are the obvious ones, like noting that computers are much faster at calculating DFTs and Fibonacci numbers than any human I know, and there are the nonobvious ones. Did you know that computers have designed some new, patentable electric circuits through evolution? That's human-competitive design "intelligence", right there.

Be careful in extrapolating this point too far, though.

To who ever said that, computers may have the ability to calculate trillions of numbers per nanosecond or crunch floating point numbers but computers are no more smarter than lets say the lowest form of life that uses a nervous system. Did you know that a single neuron has as much processing power as 5 of the worlds most advanced supercomputers?

Did you know that computers have designed some new, patentable electric circuits through evolution? That's human-competitive design "intelligence", right there.

Guess who had to program them? who had to show them how to design them? Humans right?

anybody_mcc · Post by **anybody_mcc** » 2005-09-15 01:17am

Dennis Toy wrote:
Did you know that computers have designed some new, patentable electric circuits through evolution? That's human-competitive design "intelligence", right there.
Guess who had to program them? who had to show them how to design them? Humans right?

Do you know anything about GA ? And let's say we will build intelligent being similar to humans , you will still say , there is nothing important , because we build them.
And you are quite frankly not worth mentioning as an intelligent being , because came to be through embryogenesis from the egg and this process was "developed" by evolution. So you're saing that humans have no intelligent behavior because evolution "created us" ?