Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: GMs learn program offset increased hardware speed in future tournaments

Author: Dann Corbit

Date: 10:28:24 07/26/00

Go up one level in this thread


On July 26, 2000 at 13:18:26, Jonathan Lee wrote:

>GMs will find more weaknesses through pattern recognition.
>At the same time the programmers (software) will find a way to fix them.
>In addition, hardware speed will probably increase giving GMs more on "their
>tippy toes".
>The dynamics of the ratings are rather complex.
>The question becomes when the software goes from loss to draw, and then draw to
>win.

You have pointed out the lone weakness in the analysis -- that the events are
independent.  I think clearly they are not.  The blockade idea was pretty
clearly mimicked, and a super GM is very unlikely to throw a stonewall at
another super GM so it was clearly an anticomputer strategy.  Of course, the
computer programs also learn.  Therefore, both pieces of the puzzle are moving
targets.

On the other hand, there is not a lot of evidence that this type of defence
against computers is going to be effective or that it can't be countered.  It
may also turn out that the computers learn better than the GM's do (I consider
this unlikely, but I don't know for sure).

Time will tell.  All in all, though, I think we really must admit that at least
for current performances, the ~2500 figure is certainly still reasonable.



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.