Checkers Is Solved: Perfect play by both sides leads to a draw
sciencemag.org
sciencemag.org
The guy actually seemed a bit nutty to me. The motivation for the whole endeavor was to demonstrate that his AI player (Chinook) was provably better than an amazing human checkers player by the name of Marion Tinsley. In a 45-year career, the man never lost a World Championship match, and he only ever lost 9 games (2 to Chinook). Do keep in mind that championship checkers matches are 39 games, meaning that he played thousands and thousands of games.
Anyways, a PDF of his talk is available here: http://www.cs.potsdam.edu/sigcse07/schaefferTalk.pdf . It was an interesting story, even if I thought the presenter was a bit strange.
To me, a program that already knows the right answer and simply looks it up in a database is much less interesting than a human who arrives at a similar answer through reasoning and intuition, or even an AI program that uses heuristics and some kind of internal scoring mechanism.
That's not even mentioning that I think a multi-year computation on something like solving checkers is really a (nearly) useless endeavor. Whose lives will be made better by this? Only in academia could such resources be directed at something with no tangible end value. I guess it's an accomplishment for Schaeffer, but to me it seems like anyone could throw a bunch of CPUs at a program that iterates over all possible checkers positions. I just can't imagine getting much satisfaction for such an accomplishment.
Science marches on...