Author: Jorge Pichard
Date: 07:10:22 12/25/01
The IBM team, meanwhile, has augmented the machine's hardware by adding parallel nodes, which double its effective processing speed, and by sending the software to "chess school" under the tutelage of Grandmaster Joel Benjamin. IBM also has made psychologically motivated changes--what one might call an "antihuman" strategy. For instance, the Deep Blue team has programmed the machine to prefer wide-open positions, even if they would otherwise be evaluated as slightly less promising than quieter continuation. Here the object is not so much to play perfect chess as to play in a fashion that accentuates the machine's advantage over the human. It has been reported that the machine now knows to speed up its own play when its opponent finds himself short of time. This approach, which is all too common among beginning players, often proves fatal when used by humans, for it amounts to renouncing one's advantage in time. But computers sacrifice less of their strength by playing fast than would a human would because, unlike humans, computers examine a geometrically expanding tree of possible variations. Each move deeper into the tree yields roughly the same increment in playing strength as the previous move, but it takes several times longer to complete. A human who plays in 10 seconds rather than a minute may forfeit the equivalent of 200 rating points in strength; a computer would give up perhaps a quarter as much. So far, the most prominent aspect of this struggle of alternative playing styles has been the efforts by Kasparov to close the game and by Deep Blue to open it. In Game 1, the machine managed to rip open the position with a number of violent pawn exchanges, but Kasparov countered with the sacrifice of a rook for a bishop and a pawn--approximately the material equivalent of losing a pawn. In return, he got a powerful pair of pawns deep into Deep Blue's territory. The game might still have been a fight had not the computer unaccountably exchanged queens, producing a simply won endgame for Kasparov. In a sense, Deep Blue got the kind of game it wanted but failed to capitalize on it. --------------------------------------------------------------------------------
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.