when machines begin to surpass us in intelligence

Looooong ago, my science teacher asked the class if it would be possible for a machine about half the size of a basketball to perform all the functions of the human brain. Of course, everybody said "Oh hell no, of course not". He said, are you all sure? Everybody agreed it's obviously impossible.

Then he told us: you're wrong, you've each got one in your head.

So before you pronounce that machines are forever incapable of original thought, remember that if you hadn't experienced it personally you'd swear that a gallon of goo couldn't do it either.

The brain is incredibly complex, but computers are catching up fast. It's been less than 100 years since Eniac, less than 200 since Babbage's Difference Engine. We will soon reach the limits of human-written procedural programming, and have to come up with something else like, oh, heuristic programming, learning systems and digital evolution.

Only the computers will know what's in the box. When it gets so complex that you can't understand it without spending a couple of decades tracing and dissassembling, how will you know it's not thinking for itself?
-------------------
So let's go up to the laaab and...see what's on the slaaab.

If you ask a chess playing program which is taller, a pawn or a bishop, what will the response be? Most likely it will report an error on the input of the question.

Ask a human chess player and what will the response be? Most likely wonder about the reason for the silly question.

von Neumann machines manipulate symbols, they do not understand them. The AI advocates want people to be in awe of computers but not comprehend them. A cruise missile may fly through 3 dimensional space but that does not mean it comprehends it. Why doesn't it comprehend destroying itself?

Imaginos1892 wrote:Looooong ago, my science teacher asked the class if it would be possible for a machine about half the size of a basketball to perform all the functions of the human brain. Of course, everybody said "Oh hell no, of course not". He said, are you all sure? Everybody agreed it's obviously impossible.

Then he told us: you're wrong, you've each got one in your head.

So before you pronounce that machines are forever incapable of original thought, remember that if you hadn't experienced it personally you'd swear that a gallon of goo couldn't do it either.

The brain is incredibly complex, but computers are catching up fast. It's been less than 100 years since Eniac, less than 200 since Babbage's Difference Engine. We will soon reach the limits of human-written procedural programming, and have to come up with something else like, oh, heuristic programming, learning systems and digital evolution.

Only the computers will know what's in the box. When it gets so complex that you can't understand it without spending a couple of decades tracing and dissassembling, how will you know it's not thinking for itself?
-------------------
So let's go up to the laaab and...see what's on the slaaab.

An exceptionally simple and recent real world example says very strongly that you´re wrong.

If you´re getting a new computer right now, if the prices were the same(ie no Intel price gouging), would you have any reason what so ever to pick an AMD cpu right now?

Unless you were doing bigscale mediacoding and a small number of other tasks, the answer is no.

But what if those AMD cpu:s had 20% extra performance at 20% less heat/power?

You see the current batch of AMD cpu:s became highly suboptimal because they were designed by computers. They tried to save a big chunk of money by using the same software that is used for designing graphics chips(where optimisation is drastically less important), and skip on human optimisation.

The result, is the very realistic estimate that they lost 15-25% performance and added an extra 20% of wasted power.

Unofficially they have already agreed that it was a mistake.

These are chips where components are already counted in the hundreds of millions.

umbrarchist wrote:If you ask a chess playing program which is taller, a pawn or a bishop, what will the response be? Most likely it will report an error on the input of the question.

Ask a human chess player and what will the response be? Most likely wonder about the reason for the silly question.

von Neumann machines manipulate symbols, they do not understand them. The AI advocates want people to be in awe of computers but not comprehend them. A cruise missile may fly through 3 dimensional space but that does not mean it comprehends it. Why doesn't it comprehend destroying itself?

+1.

umbrarchist wrote:von Neumann machines manipulate symbols, they do not understand them. The AI advocates want people to be in awe of computers but not comprehend them. A cruise missile may fly through 3 dimensional space but that does not mean it comprehends it. Why doesn't it comprehend destroying itself?

Because we've been very careful not to tell it about that!

What is thought? What is consciousness? How do they work? Until somebody can demonstrate that the operating principles of brains and computers are fundamentally incompatible, any pronouncements that computers can NEVER actually think are at best...premature.

Brains have a 2 billion year head start. Let's at least give computers a couple hundred before dismissing them completely.

Already we've gotten some very surprising results just by applying the Infinite Number Of Monkeys principle. Given a start, a vast field of permutations, and a way to reject inferior solutions, a computer can arrive at a correct answer that a mathematician might never find because the initial approach didn't look promising.

I agree that right now the state of A.I. is Artificial Idiocy - programming consists of giving the artificial idiot a set of exact step-by-step instructions down to the tiniest detail and if one bit is wrong, it will do something truly idiotic.

Trilobites weren't very smart either. Computers are at about that level of development now.
------------------
Postulating infinity, all else follows as a matter of course.

Imaginos1892 wrote:Brains have a 2 billion year head start. Let's at least give computers a couple hundred before dismissing them completely.

Have brains existed for 2 billion years?

200 years from now the computers may not be von Neumann machines. Or it may be possible to make them so numerous and so small that calling them that may not make much sense.

But talking about the technology we have RIGHT NOW and probably in the next 50 years being "intelligent" is most likely nonsense.

I did not say what would NEVER happen. But I would not say that it will either. I do not have a problem with waiting to see but I probably won't be here in 50 years.

Unless someone has done assembly language programming I don't get to excited about what they say regarding machine intelligence. How many registers are in a single CPU? How big are they? So what can a CPU conceptualize?

Listen to this video at 7:38

https://www.youtube.com/watch?v=aSNrKS-sCE0

"A computer is not understanding."

A computer just follows the programmed instructions.

Currently, you are correct but the day in the not to distant future is coming when computers learn from their mistakes. That's the whole point of when machines begin to surpass us in intelligence post, it's to discuss what potentially is coming. Can a machine experience emotion?

Actually, developing big brains didn't take "that" long in the grand scheme of things. Complex, multi cellular life burst onto the scene about 500 milion years ago. Large brained mammals finally got their start 65 million years ago when the Dinosaurs died out and the mammals could expand into all the empty ecological niches.

Large brained primates began evolving about 5 million years ago, and Hominids developed @ 2 million years ago. Anatomically modern humans are about 200,000 years old, and what can be recognized as modern human behaviour dates back 50,000 years.

So if computers have evolved to the level of Trilobites, then we only have to wait a while...

smr wrote:That's the whole point of when machines begin to surpass us in intelligence post, it's to discuss what potentially is coming. Can a machine experience emotion?

That is an ASSUMPTION or possibly a SPECULATION.

The machines can store and manipulate symbols faster and better then we can already.

The problem is UNDERSTANDING the symbols. I am not saying it can't happen. But just saying more powerful von Neumann machines will do it when most people don't understand von Neumann machines does not cut it for me.

The Two Faces of Tomorrow is the best AI story I know of.

Are emotions INTELLIGENCE? Most AI stories just have neurotic human minds in a box.