Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Deep Fritz - How many games does it take before it "Learns" something ?

Author: Dann Corbit

Date: 15:47:17 10/09/02

Go up one level in this thread


On October 09, 2002 at 18:08:44, John Wentworth2 wrote:

>I have never tested the learning features in the various programs. If they lose
>will they actually change their moves if you play exactly the same way again?
>
>If Kramnik is beating Deep Fritz because of it's style of play would it not
>"learn" to change what it's doing?

There are a lot of pathways in the game of chess [cough: Deliberate
understatement].

Book learning won't help.  TD-Lambda learning slows actual play (though it could
likely digest a thorough log afterwards).

Don't forget, Kramnik is learning as he plays too.  I suspect his learning
function is pretty advanced.

I don't think this changes anything about the "Are computers GM's?" debate.

Kramnik is several standard deviations above the norm, even for a GM.

The truly elite players are really superstars at chess.  Like an 'A' player
tangling with GM's, GM's tangling with super GM's is qute a tussle for the
ordinary GM's.

If Fritz could win a match like this, it would be astonishing.

Eventually, the computers will have their day.  It's inevitable.  But (for now)
the most elite chess players are pretty amazing.

At some point, even correspondence GM's will have to succumb.  That's why this
point in history is interesting.  The best programs and the best people are of
similar ability.



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.