Re: Aaron Sloman's "The Irrelevance of Turing Machines to AI" article

From: Wolf Kirchmeir (wwolfkir_at_sympatico.ca)
Date: 08/04/04 

Date: Wed, 04 Aug 2004 09:35:20 -0400

Sergio Navega wrote:

> "Wolf Kirchmeir" <wwolfkir@sympatico.ca> escreveu na mensagem
> news:FEMPc.15422$Jq2.711179@news20.bellglobal.com...
>
>>Neil W Rickert wrote:
>>[...]
>>
>>>My own view is that, in some sense, learning is adaptation. And I
>>>think computers can only adapt in ways that they have been
>>>preprogrammed to adapt. They cannot deal will with the unexpected.
>>
>>Careful. As soon as you say a program can't do something, someone will
>>come along and write a program that can do it.
>>
>
>
> Although I'm at your side here, I must admit that we do have some
> things that *cannot* be done by computers. One of them is a program
> able to tell if a generic program will stop or not. But that's a
> mathematical question, not a real world one (besides, neither a
> human being is able to do that).
>
> Sergio Navega.

Precissly.

Most of the time, when people claim that computers can't do something,
they refer to some human ability. I recall the time not so long ago when
it was argued that no program could play chess at the master level, let
alone grand-master. Hah!

It should be obvious to anyone with a bit of cross-displinary knowledge
that a computer is in principle capable of performing any computable
task. It should also be obvious that we don't know the computability of
a large number of "real world" taks, so that claims that a computer
can't do some one of them are usually expressions anthropocentric
prejudice (and vanity, too, porbbaly.)

Besides, many non-computable tasks can nevertheless be done to some
acceptable level of "good enough." Which may even be the best (ie,
"correct") solution, but we'll never know for sure.