Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: 5 GMs, 5 computers, 3 months

Author: Russell Reagan

Date: 12:36:07 04/13/02

Go up one level in this thread


This is a tough topic, but I'll give it my best shot.

I think there are two things we need to consider.

First, our top human players are not spending their time studying how computers
play chess. Some might spend time analyzing computer games if they have an
upcoming match against a computer, but as a whole, they don't study how
computers play chess and while they do have limited strategies against
computers, the full weaknesses in a computer's style of play may be a domain
that has barely been explored at all.

Having established that humans aren't really focused on rooting out computer's
weaknesses, we can't really determine FOR SURE if our best human players would
be able to find weaknesses in the computer's play. If you think about it, you
realize that finding weaknesses in their opponent's play is precisely what
modern GM's do, so I would hypothesize that if GM's decided to really attack the
problem of finding computer's weaknesses, they would find some nice methods of
attacking computer chess programs in future matches.

The second thing that I consider to be of interest in this discussion is that it
seems highly likely that a perfectly played game of chess will result in a draw.
If that's not the case, the eventually computers will beat humans every single
game. However, since it is likely that a perfectly played game of chess will
result in a draw, it is not impossible that a time will come when computer can
play perfect chess and humans in turn (learning from the computers) can play
perfect chess in response, and every game will end in a draw. That time could
also be the end of competitive chess, because if there are a handful of humans
who are literally invincible on the chess board, they will draw with each other
and demolish everyone else. It will merely be a matter of striving for the goal
of becomming one of those perfect chess players.

I personally don't see this happening, but there are people in the world much
smarter and much more creative than I, so I would not rule it out for a select
few.

My opinion is that right now, GM's could compete with the top computer programs
if they chose to do as you say and dedicate significant time to studying the
programs, but there is no real incentive to do so. Unless someone wants to pay
money to have this experiment carried out, I don't see why any GM would want to
do it. Think about it. If the GM's gather together and analyze the computer's
style of play and find all kinds of weaknesses, their work will be meaningless
within a few years. The programmers will develop techniques of patching the
weaknesses, and hardware will get faster and eliminate some of the weaknesses by
sheer brute force search. So unless a GM is just really hurting for money, I
don't see any motivation for anyone to take part in such an experiment.

The bottom line is that eventually computers will play perfect chess and never
make a mistake, ever. We know that for sure (although it might not be in any of
our lifetimes). On the human side, we do not know if humans will ever be able to
play perfect chess. It's very likely we won't be able to.

So the real area of discussion lies in the question, do you think that humans
will ever be able to play perfect chess? I don't think so, and when computer are
able to, so much for chess :)

Maybe we will move on to a "new chess" played on a larger board, or a N x N size
board where there are no edges. Or 3D chess. Or maybe real time chess where
there are no "turns" and each player simply plays as fast as he can, so the more
time you spend searching, the more time you waste. That would be interesting to
see IMO :)

Russell



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.