Author: Pete R.
Date: 09:52:48 07/26/00
Go up one level in this thread
On July 26, 2000 at 01:44:57, blass uri wrote: >I think that if you give GM's many position and ask them to evaluate what is the >expected result for white after 1 second and you give computers the same >positions and ask them to evaluate the expected result for white by translating >the number of their static evaluation function to the expected result then you >are going to find that the evaluation of computers is better because humans >cannnot calculate in one second all the things that computers calculate in one >evaluation. Well of course but this is a silly restriction. We know humans can't compete at that speed. I was assuming we let the GMs assess the position as long as they like, e.g. pawn structure, good and bad minor pieces, space, etc. etc. and then say which side is better. If a program's evaluation function produced the same results, even with occasional misevaluations, it would essentially be the equivalent of a human GM who can flawlessly visualize future positions to great depths. Kramnick's play against DJ would have been impossible against such a program, as would any other type of anti-computer play. Tom is asking for a test suite of positions that have such a GM concensus evaluation. It's a difficult prospect to create such a thing. Perhaps one could create a test suite based on positions that computers have already blown against humans, an anti-computer/closed position test suite. For open positions program tactical evaluations seem to compensate well enough for gaps in positional knowledge.
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.