Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Once GM's figure it out, it's all over (for a while)

Author: Robert Hyatt

Date: 07:26:58 01/06/00

Go up one level in this thread


On January 05, 2000 at 16:36:22, Dann Corbit wrote:

>On January 05, 2000 at 16:09:41, Chris Carson wrote:
>[snip]
>>I agree with you.  This is a weakness of the programs.  If a GM is prepared
>>for a program and knows the programs weakness, it is over.  Although the
>>same maybe true for a human opponent of a GM (hmm?).  :)
>>
>>By compete, do you mean win more than loose to a 2500 FIDE GM?  Or is
>>it something different?  I am not sure what mean what people mean when they
>>say compete.  I am not sure what they mean by GM strength.
>
>By compete, I mean really be at that level.  The computer programs are slowly
>gaining positional knowlege.  If you look a the evals for Phalanx or Crafty, you
>will see positional understanding.  It is even better for some of the commercial
>programs.  I think that a computer program can be made to recognize that it is
>in a locked position.  In such circumstances, they may change strategy.
>
>>Rebel century is close to 2500 (between 2450 and 2500) performance rating
>>at match play.  My opinion: The rating would be higher on an Athalon 800.
>>I also think tournaments provide a better rating estimate than match play
>>(for the reasons I stated above about preparedness).  :)
>The GM's that play against them have been playing normal GM type games.  See
>what happens when they play D00 or some other locked formation.  Let them study
>anti-computer strategy and play 100 games against the target program.  I think
>the rating would drop at least one hundred points.
>
>>When I play, I consider anyone within 200 points competative.  They have
>>a good chance (not above 50%, but not zero) to win.  Am I off base here?  :)
>
>With improvements in hardware and software, in four years computers would be
>able to complete a GM norm, even with:
>1.  GM's playing against the machine as often as they like to learn the
>strengths and weaknesses


I disagree there.  If you let a GM play a computer over and over, and _then_
play the same program in some serious games, the computer is going to catch
hell.  Humans evolve.  Computer programs do not.  Once he finds a hole, he
will exploit it over and over, while a human would 'learn'.

That will be a weakness for another 25 years or more.




>2.  Closed and positional strategies plied against the programs
>
>I think that is true for two reasons.  First, the programs will be 100 ELO
>stronger just from hardware.  Second, the programs will gain enough positional
>knowlege and stragegic understanding to compete, though GM's will still be
>better.  However, the ability to see any tactic that is 7 moves away without
>fail will compensate for the lesser strategic ability, even with anti-computer
>tactics.
>
>I'm no Nostradamus, but I think the prediction will hold true.
>It's really more of a guess, though.  Let's call it that.



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.