No, there was no official championship for this game, but this time the algorithm managed to surround Ke Jie, the world's best player in go, by as much as half a point.
The head of Google Deep Mind Demis Hassabis explained that this time the AlphaGo was programmed in such a way that it would increase its chances of winning already at the beginning of the game, but the program settings presuppose other options for AI work. You can, for example, program it to increase the gap on points.
During the competition Ke Jie several times in the match ahead of the program, but AlphaGo after a while again began to outplay the champion, because AI was originally programmed to win. Perhaps, that's why the difference in the account turned out to be quite small. If the program were tuned differently, it could start winning additional points, increasing its indicators by the end of the game.
In the near future there will be two more competitions in the game of Go, in which, in addition to people, artificial intelligence will take part. It is known that during one of the matches the program will fight at once with five players.
The new version of AlphaGo is significantly improved and compared to the one that recently won 60 players on Chinese websites, it requires almost ten times less computing power.
The article is based on materials .
Comments
Post a Comment