2008-08-08

Computer go landmark.

At the US Go Congress, played at the same time as the European Go Congress, they had a computer go event where a 8 dan professional lost to a computer program. OK, I think the program got a 9 stone handicap on a 19x19 board. The article in American Go Associations e-journal is a little bit cryptic. But anyway, it was a program based on the Monte Carlo algorithm, which would scale well with parallel processing. The program used 800 processors borrowed from a supercomputer. It was also featured on slashdot.

Monte Carlo and the brute force used for chess give strength to the programs. But I would not call it thinking. I wonder if soon we could study some new and strong moves made by programs like this to improve our thinking.

No comments: