(Kazan, January 29, “Tatar-inform”). Scientists have created a computer program that was able to defeat a professional Go players. This was reported in the journal «Nature», writes Lenta.ru.
The logic board game first, which arose in ancient China, is considered the “hard nut” for artificial intelligence – because of the large size of the board and the difficulties Estimates of moves.
Play two: one gets black stones, the other – white. The purpose of rivals – isolate on the board with stones of the same color more territory than the opponent. The best of the program is to reach a level of amateur players.
The international team of scientists has presented a program called AlphaGo. It uses the so-called “network value” to assess the pieces on the board, and “network rules” to select the moves. These neural networks learn to play by analyzing the known parties, and through trial and error, playing alone.
AlphaGo beat other programs to 99.8 percent batches, and then achieved superiority over the European champion in the game of go (with a score of five to zero), without any handicap. Now AlphaGo waiting for the tournament with a higher level of pro – Korean Lee Sedol (second in the world in terms of international titles).
*** Sf
If you find a mistake, highlight text and press Ctrl + Enter
No comments:
Post a Comment