08 December 2006

The Machine Sees Further Ahead

The recent Kramnik - Fritz match resulted in an indisputable victory for the machine. This set it apart from the previous match won by a computer over a World Champion -- the 1997 Kasparov vs. Deep Blue match -- where Kasparov's unfounded accusations of cheating convinced the more gullible members of the non-chess playing public that foul play was involved.

A decade ago chess playing software passed from not-as-good as a person to at-least-as-good. Since then the software, coupled with better hardware, has passed to better-than. What changed? The machines can see even further ahead than before. A paragraph in David Shenk's 'The Immortal Game' drove this point home for me.

With so much fluidity in the game -- a near-infinite number of ways to win and a near-infinite number of ways to lose -- a newcomer might reasonably assume (as I certainly did) that chess is mostly a game of quick thinking. Since a game is won or lost on a player's ability to outmaneuver an opponent's pieces, and since it is surely impossible to memorize or analyze even a tiny fraction of all the possible board configurations, one would naturally expect most games to go to the sharpest -- or deepest -- thinker, the player able to see the furthest ahead. (p.77)

While other factors might tip the balance for person vs. person, or for machine vs. machine, in a match between a person and a machine the computer is unquestionably the player 'able to see the furthest ahead'. The grandmaster's better understanding of chess no longer compensates for the machine's advantage in calculating variations. For man - machine matches to become fair fights again, some other form of compensation is required.

Can Kramnik beat Fritz if he starts with an extra Knight? With an extra Pawn? With equal material but two consecutive moves at the beginning? There is only one way to find out.

No comments: