AI, at least as it was originally conceived, was not and is not being used in the more successful attempts to make computers play either Go or Chess well. As unsexy as that is, it's just the facts. AI has failed badly in these tasks, and there's not much to be done about it. Something akin to machine learning techniques are being used in some Go programs, but the most successful Chess programs are essentially board evaluators, and (many of) the most successful Go programs are play-line samplers.
The reason Go is harder has nothing to do with anything that RAND or Deleuze might want to think, either. The board is bigger, the number of possible games is bigger, and although at a quick glance, you can see if there's a large material difference between white and black in chess by simply adding up the value of the pieces*, looking at player-to-move and a few other board characteristics, situations which are relatively settled on the board in Go between strong players might be completely unsettled to a computer because a computer can't quickly estimate the value of the board in Go until the game is nearly complete. So two main (oversimplified) reasons: 1) any given chess board position is relatively easier to evaluate, and 2) there are many more possible Go games than chess games. Scifi, critical theory, and political theory notwithstanding, it essentially comes down to the arithmetic behind the games. s. * yes, modern chess evaluators can be more sophisticated than this, but they don't really need to be to be successful.
_______________________________________________ Computer-go mailing list [email protected] http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
