At Tuesday's meeting, Abram described his work on the Google AI challenge. Here, the winner of the challenge explains his program in some detail:

http://a1k0n.net/blah/archives/2010/03/index.html

I remember reading about the Netflix prize, in which winners tended to apply several different algorithms to the problem, selecting between them in different situations, rather than optimizing the hell out of one fundamental algorithm. Seems similar to this case.

- vish