Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Artificial Intelligence

Author: Bruce Moreland

Date: 22:47:39 01/27/98

Go up one level in this thread



On January 28, 1998 at 00:41:26, Don Dailey wrote:

>But the real problem with these discussions (in my opinion) is that
>there is no definition of intelligence we agree on.  Everyone THINKS
>they know the answer intuitively but I'll bet everyone has a different
>idea (highly internalized in their heads) about what it really is.

This is the cause of all of the randomness, but it is a hard topic.
Even the definition in the OED is pretty lame (a whole page of it).  At
least I think so.  I was trying to read it without using a manifying
glass.

This whole mess started over an article that discusses the issue of
whether or not Deep Blue, and by extension, any normal chess program,
uses techniques that could properly be considered to be in the domain of
artificial intelligence.

The author of the article says yes, IBM's web site says no.

What caused the real problem was the KK's assertion that the author of
the article said that Deep Blue is intelligent.  This assertion was not
true, but by then we were all off to the races.

This kind of thing is what the internet is all about.

I am not a big Carl Sagan fan, but in one of his books, I believe it was
The Dragons of Eden, I think he suggested that intelligence is an
individual's capacity for adaptation.  Sorry for the vagueness but I
read all of his stuff in approximately 1980.

Personally, I think this definition is a good one, better than the one
in the OED at describing this part of the word's meaning.

I think this definition is a good reference point when you try to figure
out what artificial intelligence is.  One of the posters in this thread
has mentioned that he thinks that AI has to do with solving problems
that require a lot of generalization.

I agree with this, I had come to this conclusion independently.

The reason chess was interesting to AI researchers is that it is hard to
understand any one position, and there are a huge number of positions
that need to be understood in order to play the game *well*.  A program
that plays chess has to be very adaptable.

In order to attack this problem people used min-max search and end-point
evaluation.  This is an attempt to apply a generalized heuristic
approach to the game of chess.

Problem is that it worked, and improvements have come for free along
with improvements in hardware capacity, so people haven't had to resort
to anything fancier in order to play a strong game.

The reason AI people are bummed about computer chess is that the systems
that perform best are not generalized enough to interest them.  More
generalized systems can't out-peform the more specific chess programs as
of yet.

So I agree with the author of the article.  This stuff is AI, but it's
not cutting-edge AI, although it's all cutting-edge computer chess of
course.

bruce



This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.