Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: electric GM already here super gm 5 years

Author: Christopher R. Dorr

Date: 09:25:23 12/12/00

Go up one level in this thread


On December 12, 2000 at 10:11:00, walter irvin wrote:

>pc programs can easily hold their own vs gm's .super GM or the top ten are only
>5 years away from being over taken .but i dont think that is the end of it .in
>25 years the game of chess will be solved (e4 and mate in 95 ).e4 nc6 ??? oh
>mistake now mate in 67 .computers are still getting better people are not
>.people can not say  im upgrade my brain to 600 ghz .i believe this time table
>would be much   much sooner if there were more money in it .would be cut to 5
>years easy .

Right now, computers can generally hold their own with GMs in *limited*
contests. Other than Deep Blue (which clearly was solid GM Strength), just about
all current computers are solid IM, perhaps weak GM strength, over the long
haul. If you took Fritz6, for example, and put it out on the real GM tour, over
the course of a year or two, it would likely show itself to be weak GM strength
at 40/2, as GMs were forced to take it seriously as a regular opponent, and
learned it's strengths and weaknesses. What motivation does a GM have to develop
techniques to bust computers right now? He won't play it in a bread-and-butter
tournament. Shouldn't he spend his time preparing for GM Randomov, who he will
play 5 or 6 times this year, with real money on the line?

As Bob Hyatt said, comps lose all the time on ICC to good human players. At
blitz. At 40/2, their results would be *much* worse.

As to chess being solved, and a computer saying something like 'mate in 67'
outside a very limited tablebase, do you have *any* idea how much processing
power and storage it would take to do something like that? It's not even
conceivable that that could take place withing many decades. The numbers are
astounding. I remember seeing a very rough calculation demonstrating that to
completely solve chess within 100 years would take a program able to compute
(very approximately) 10^45 nodes per second. Please correct me if I'm incorrect.
I just vaguely recall seeing it. But the exact number isn't important. There is
no *theoretical* way to get something like that. You are assuming that some kind
of breathtaking advance in computer science will take place before then. It may,
but why on earth would you count on it? I'd put much more money on the
probability of pigs evolving wings than on chess being solved within the next 50
years.

Chris




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.