Author: Walter Faxon
Date: 23:58:07 01/07/03
Go up one level in this thread
On January 07, 2003 at 05:51:26, Graham Laight wrote: >Hi Walter, > >Impressive answers! More comments by me below. > >On January 07, 2003 at 04:04:39, Walter Faxon wrote: <snip> >>(1) If we want a venue to explore advanced AI in chess, let's change the rules >>for a new series of competitions: "limited search" computer chess (LSCC). >>Humans search about 2 positions per second (Anand claims he searches 5). Well, >>our computers today are pretty dumb so let's start with a maximum 100 >>positions/second; this can be lowered later. The computer horsepower now used >>for searching would be applied to complex pattern recognition and other AI >>techniques. All program sources would be made public after each tournament, >>like the "RoboCup" robot soccer competitions today. This would serve two >>purposes: to prove nobody is searching faster than allowed, and to spread the >>wealth of knowledge. This would revitalize academic computer chess: everybody >>learns from everyone else and there is no point in competing unless you have a >>new idea or a much better implementation. Many more papers would be written. >>And any ideas developed could also be adopted by the fast searchers, as >>possible. > >I would LOVE for this competition to happen. If enough people were willing to do >this (which, regretfully, is unlikely), then I'd happily take part! Someone has to go first; why not you? Write your program, making it as general as possible with sockets for various AI methods, do experiments, write a paper or two, then make it all public. Others can plug in their own ideas and methods, and boom! we have a small herd of pattern-oriented limited-search chess programs with authors eager to show them off. For one pattern-based approach to chess, you might revisit the 1979 PhD thesis of David E. Wilkins (See http://www.ai.sri.com/~wilkins/bib-chess.html). An impressive work, and the basis for a couple of papers in the AI Journal. A possible start would be to reimplement his logic in Prolog. >>(2) The best over-the-board chessplayers make a _lot_ of mistakes, particularly >>allowing small advantages to dissipate. The much lower rate of draws in >>top-level correspondence chess suggests that very minor advantages _can_ be >>exploited, given enough time, energy and imagination. A super-grandmaster > >It might suggest it, but it falls far short of proving it. If you go to the SSDF >website ( http://w1.859.telia.com/~u85924109/ssdf/ ) and look at the games, >you'll see that games between stronger chess programs are more likely to end in >draws than games between weaker ones. This, together with the fact that it has >been proven that there's no forced win of material in the first 30 moves, proves >beyond reasonable doubt that chess is a draw with optimal play. I disagree. Stronger computers draw each other more often because each can avoid falling victim to the other's tactics. They are not exploiting subtle advantages; they don't know how. They stumble around a while and if neither side happens into the winning line that neither side sees, they draw. Why else is a computer plus a competent human player (not necessarily a grandmaster or even a master) so much stronger than a computer alone? Even if there is no forced win of material in the first 30 moves (determined by a full 60 ply search!?!?), that doesn't mean that one side doesn't have a winning advantage. There are books of openings analysis where (for example) a variation is refuted by a line where one side queens one move ahead of the opponent, on move 75. Do you have a URL for this "no forced win of material" claim? <snip> >results show it! The very worst Warri programs will EASILY beat the human world >champion, wheras in Go, I'm told that even human beginners can beat the very >best programs. <snip> >-g My understanding is that the best Go programs are weak amateurs, still much stronger than beginners. But I'm not sure of that; I'm not that interested in Go. Because we don't need Go for a research platform. Every AI problem that is found in Go is also found in chess and we can force ourselves to confront these problems if we throw away the crutch of brute-force search! -- Walter
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.