Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Artificial Intelligence

Author: Robert Hyatt

Date: 18:49:53 01/26/98

Go up one level in this thread


On January 26, 1998 at 17:59:04, Dan Homan wrote:

>I didn't mean to offend, I'm just trying to figure out what you
>mean by "artificially intelligent".  I question your definitions,
>and you seem to fall back to the position that "chess requires
>intelligence, therefore...."

I don't know whether this is true or not... and really have no way to
argue for or against.  It is simply " a fact in evidence" from
statements
made in nearly every AI book I have read.  I don't even want to try to
define "intelligence".  My only interest is in working on a computer
chess program and making it better.  I belive, based on prior comments
by many AI researchers, that chess is a game that requires intelligence
to play.  If that's false, there's plenty of blame to go around as this
has
certainly been mentioned many times.  I assume it to be true, based on
the
reputations of those that made the statement and based on no evidence to
the contrary...

If the computer can play a game that requires intelligence to play, then
obviously the computer has to have some sort of "artificial
intelligence"
component to play the game.  Else the entire premise is false.  If you
want
to make a case that chess doesn't take intelligence, feel free.  That I
can't
discuss as I'm not into that sort of stuff...



>
>I would present an opposing view...  that the success of computer
>algorithms at chess has demonstrated that chess does not require
>intelligence.  Simply following a set of rules over and over again
>can produce outstanding results previously believed to be only
>achievable via human intellect.

The problem might be, *what* would you give as an example of something
that can't be solved by such an approach?  IE what if we one day
discover
that all of our mental processes are really simple 0/1 operations?


>
>I see your point, however, that computer programmers have succeeded
>in artifically solving a problem which was previously *assumed* to
>require human intelligence.  In that sense I suppose they are
>"artifically intelligent".  But I reiterate my concern that this
>is a very misleading term.  I like your suggestion of "artifical
>reasoning".
>
>Rather than bicker over the definitions we are carrying around in
>our heads, maybe we should lay out what we mean by intelligence...
>
>1) Ability to solve complex problems
>2) Ability to generalize from past experience to new experience
>3) Ability to learn something completely new
>    (something beyond the intentions of original programming)
>4) Ability to create
>
>I'm sure there are other things that people would like to include,
>but lets start with these.  First look at a basic brute force
>chess program with no learning capability....
>
>Such a program plays chess very well, satisfying the Turing test
>with your modification of limited domain, but it will only have
>the first and fourth criteria here.  (I am assuming that finding
>new opening lines and new solutions in endgames can be considered
>a limited form of creativity.) A program that learns its
>evaluation function will satisfy the second critera as well.
>That's 3 critera satisfied to some degree out of 4 which isn't bad.
>
>The interesting thing about Turing's original test (with no
>restriction on the kind of questions that can be asked) is
>that a program would need to satisfy all four critera to pass.
>An program that didn't would be unveiled over time.  That is
>why I thought that restricting the domain of the test undermined
>it as an "intelligence" test.
>
>The more I think about it; the more I come to the opinion that
>it is generality (critera 2 and 3) that makes intelligence
>so powerful as a problem solving device.
>
> - Dan



This page took 0.04 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.