Author: martin fierz
Date: 16:09:14 02/21/02
Go up one level in this thread
On February 21, 2002 at 18:23:30, Otello Gnaramori wrote: >On February 21, 2002 at 17:18:52, martin fierz wrote: > >>>i would say, computers today resemble a manic-depressive supergrandmaster in >>their results - not in their play. there are many positions which they do not >>understand. they can achieve the same performance as a grandmaster quite often, >>and yet they are totally clueless in whole classes of positions, > >Very few indeed nowadays and especially found in endgames, but there are >tablebases for that... no. not very few. everything closed. everything involving trapped pieces too. dann corbit recently posted a position asking "which program can see that this position is dead lost?". i think none could. there was not much response to that post by all the guys with the fast machines and the good programs :-) i could see it easily - with my lousy 2200 rating. the concept of "this piece will never move again" is not in the programs AFAIK. i have written one of the best checkers programs around, and it is grandmaster-like in performance without doubt. yet i know that it has serious weaknesses when trapped pieces are involved. i don't know how to explain the concept to the program :-(. i can (and have) put in lots of special cases - but a general "understanding" which does not rely on thousands of "if then else" rules would be much better. chess, of course, is much more complicated than checkers. tablebases only work with very few pieces, and it is unlikely that we will see larger tablebases than 6 pieces in our lifetime (i claim - is it true?). > >which even >>human beginners can understand. >>people like me say that computers are not grandmasters, because they have >>completely different capabilities than a grandmaster - they reach the same ends >>by *totally* different means. i'm not saying they play worse performancewise. >>but they are something different. you cannot compare apples and oranges :-) >> > >If that is your point the match between Human and comps are nonsense for you ? not at all: they are very interesting. humans and computers use totally different approaches to get to the same result, play chess well. it's interesting to see how they try to lure each other on their "home ground". you can see that LvW is playing the french defense against rebel, something which i think (?) he doesn't play often; in the hope of keeping the position closed. i believe that the game of chess between humans and computers is not decided by one side playing better chess than the other, but rather by which side succeeds to get his type of play. closed position, rebel loses. tactics, LvW loses. so it's a kind of "meta-game", where you aim at a certain type of position. i like to watch the meta-game. LvW obviously knows what he is trying to do (unlike the chess tiger in argentine games, where IMO the opponents were not well prepared), but he is not succeeding as black. up to now, i would have thought that a well-prepared grandmaster (prepared at meta-game strategy, not chess-wise) would still beat a computer. but it looks like i might be wrong :-) aloha martin
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.