Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: A thoughtful answer to the tired question, "Are computers GM strength?"

Author: Graham Laight

Date: 02:40:33 07/15/00

Go up one level in this thread


I agree that this is a thoughtful answer.

However, there is potentially a paradox.

A computer which is a GM one year may not be a GM next year - even if it still
plays the same moves in the same positions - because it may have a fundamental
weakness that cannot be overcome by tweaking. In this case, other chess players
will learn this weakness and start consistently beating it.

The same could also apply to humans - though probably to a lesser extent.

This might sound very hypothetical, but I'm sure it has actually happened in the
past.

-g

On July 15, 2000 at 04:31:05, Jeff Anderson wrote:

>For a long time I have observed the debate in this forum, "Are Computers GM
>strength in tournament time controls?" or some such thing.  This is question is
>impossible to answer unless we define what is meant by "GM Strength".  To answer
>the question do we disect the programs play or the programs results?  I assert
>that the only way to aswer the question is through the results of the top
>programs.
>
>Some critics of the belief that the top computer chess programs are GM strength,
>such as Robert Hyatt, argue that 'the computer makes this move that no GM or
>even a class player would make', or that 'the computer lost in a fashion which I
>very rarely see a GM lose in', or that the computer 'lacked a plan' or
>'demonstrated a clear lack of chess understanding'.  But these arguments are
>silly.  It is true that in some positions the computer will play weaker than in
>others, but this is just its style.  If these defects were serious enough that
>the program would not be considered GM strength, then the porgrams opponents
>would prevent it from demonstrating GM results.  Despite errors in certain types
>of positions the computers have demonstrated a clear ability to demonstrate GM
>results.
>
>The computers are GM strength.
>A top program on good hardware will not perform below GM level in a standard
>time control tournament of humans ever again.
>
>Jeff



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.