Is anyone aware of any interesting research on developing strong
chess-playing engines using genetic algorithms? According to Wikipedia,
this has been done for checkers, resulting in a fairly strong program (but
not as strong as the best, Chinook). The internet mentions some efforts at
doing so for chess, but I haven't heard of any huge achievement. GA are
quite fascinating; the resulting problem solvers (for biological research,
electronic design, financial markets analysis, etc.) often end up being
something the human designer doesn't even understand, yet perform better
than what the human would have produced. For chess, it seems like one could
start with a bunch of random-playing algorithms (e.g., Stanley the Monkey in
Chessmaster!), randomly mutate them, select for fitness (i.e., chess
strength), repeat, ... After lots of generations maybe there'd be some that
could beat today's best. Interesting possibility?

DB