AI in 12 years?

[mszlazak]

Originally Posted by David Bailey

We may indeed by talking about different things, I am not quite sure.

Well lets assume the brain-as-receiver idea is approximately correct. In that case, there must be bits of the brain that do the receiving, and they would not simulate - just as if you simulated a radio, you wouldn't get the channel it was tuned to.

David

Yes! The problem is still is whether this grown physical replica is fully like a human brain and what this has to do with the human mind, consciousness and volition. That is what I initially said.

[Paul C. Anagnostopoulos]

Originally Posted by David Bailey

So why not include every bit of clever software - you can't broaden the definition of AItoo much!

A lot of clever software had its roots in AI research. You're not suggesting we narrow the field to human brain simulation, right?

Everything with you seems to amount to semantics! I thought perhaps you would buy into my old assumption that understanding was just a complex collection of procedures. How do you want to understand understanding (!!), or doesn't it exist!

Of course I think understanding exists. But some people appear to believe that it can only be done by humans, and then in some way they cannot explain.

~~ Paul

[Paul C. Anagnostopoulos]

Originally Posted by 4vektor

i think humans will eventually prove that it does not halt. this is just a particular halting problem. it is known that a turing machine cannot solve general halting problem. there is no statement that humans cannot. and maybe the point is: such a statement is not provable. (i am not saying the preceding statement is proven, hence the maybe.)

Agreed. It is an open question whether humans can solve the general halting problem. But there will always be math problems whose solvability is unknown, which means there will always be equivalent programs whose halting status is unknown.

as i understand it, the weakness in penrose's interpretation of godel is essentially the validity of this argument: humans, bc we are capable of producing godel-type results (for example proof to godel incompleteness theorem itself), cannot be algorithms. we can distinguish between the truth and provability of a statement whereas an algorithm cannot. i think i agree with this. critiques usually involve attack on godel's assumptions applicability to humans -- we do not know that that we are sound and consistent (or in fact we are not, so we cannot know a falsehood).

Why do you think that a computer could not have come up with Godel's theorems?

~~ Paul

[David Bailey]

Originally Posted by Paul C. Anagnostopoulos

A lot of clever software had its roots in AI research. You're not suggesting we narrow the field to human brain simulation, right?

Well yes, it did, it seemed to invent or popularise ways to store arbitrarily sized structures using memory allocation and pointers. It was also responsibe for some clever languages - though a lot of them have sunk into niche use by now. However, that really isn't the point!

Of course I think understanding exists. But some people appear to believe that it can only be done by humans, and then in some way they cannot explain.

~~ Paul

Sometimes it is a scientific advance to recognise that something can't be done - solving the quintic by radicals, explaining gravity by contact forces, etc. There is a danger that science ends up asserting and believing various things must be true, without ever actually demonstrating that they are true.

David

[Carl Jung]

Originally Posted by Paul C. Anagnostopoulos

What do you mean by "in a sensory manner"?

~~ Paul

The difference between third-person view and first-person ontology.
Having a relation to the world with a history, pains, fears and hopes.
Intentional states.

[Paul C. Anagnostopoulos]

Originally Posted by Carl Jung

The difference between third-person view and first-person ontology.
Having a relation to the world with a history, pains, fears and hopes.
Intentional states.

I believe a computer can have a perfectly good first-person life. But we'll have to wait to find out.

~~ Paul

[4vektor]

Originally Posted by Paul C. Anagnostopoulos

Why do you think that a computer could not have come up with Godel's theorems?

the incompleteness theorem involves the recognition of a statement that is not provable from the axioms. if an algorithm knows only the natural numbers and the axioms, it cannot understand how to find such statement (which involves a diagonal or self-referential trick). essentially it can only numerically process the axioms, will not halt, and the diagonal sequence will never be found. that is what i think.

Agreed. It is an open question whether humans can solve the general halting problem. But there will always be math problems whose solvability is unknown, which means there will always be equivalent programs whose halting status is unknown.

sure. but i was saying that if it can be proven that the statement "humans cannot solve general halting problem" is unprovable, that would distinguish humans from algorithms. i certainly will not be the one to do it and it may not be possible.

[Carl Jung]

Originally Posted by Paul C. Anagnostopoulos

I believe a computer can have a perfectly good first-person life. But we'll have to wait to find out.

~~ Paul

I do not believe computers as symbolic manipulators are any different than an abacus. I can't see how computers could ever understand metaphor and how it could be self-referential. Many speak of programs being self-referential but they do that much in the same way an algorithm can be referenced to a set of coordinates. And that just proves they haven't got the point

[Paul C. Anagnostopoulos]

Originally Posted by 4vektor

the incompleteness theorem involves the recognition of a statement that is not provable from the axioms. if an algorithm knows only the natural numbers and the axioms, it cannot understand how to find such statement (which involves a diagonal or self-referential trick). essentially it can only numerically process the axioms, will not halt, and the diagonal sequence will never be found. that is what i think.

Right, but why is the computer restricted to the natural numbers and axioms?

sure. but i was saying that if it can be proven that the statement "humans cannot solve general halting problem" is unprovable, that would distinguish humans from algorithms. i certainly will not be the one to do it and it may not be possible.

I'm not sure why a computer cannot prove that certain theorems are undecidable.

~~ Paul

[Paul C. Anagnostopoulos]

Originally Posted by Carl Jung

I do not believe computers as symbolic manipulators are any different than an abacus. I can't see how computers could ever understand metaphor and how it could be self-referential.

But your inability to see is not a proof.

We don't have much idea how humans understand metaphor nor how our memories are organized. I think it's premature to assume there's a trick there that computers could not utilize.

~~ Paul