Google secretly taught artificial intelligence in fights with people

Recently the Internet has intensified the “mysterious master” who has won hundreds of top players. Google employees have admitted that under his mask was hiding artificial intelligence AlphaGo.

How does AlphaGo

In AlphaGo uses two neural networks. The first reviews the position on the Board, the second uses the results of the analysis to determine the best move.

Artificial intelligence has developed in 2014, the division of DeepMind by Google. For testing AlphaGo the developers have arranged a match with the champion of Europe on th Fagnou Hui, who lost all five matches. In March 2016 AlphaGo won and world champion Guo Li Sedalia – he was able to win only one of the five parties.

Google starts and wins

A new strong player under the nickname Master(P) appeared in the popular systems for the go game Tygem and FoxGo. In a short time he won over many players.

Achieve Master(P) have been actively discussing on the forums, and no one knew where there was a daring newcomer. When a mysterious user known lost master. GU Lee, he promised to pay any player who wins Master(P), 100 thousand yuan (about 860 thousand rubles).

But Master(P) continued a series of victories. Of the 50 games in a draw for technical reasons, ended only one. Players began to suspect that they are played by the computer. Soon Demis, Hassabis, one of the founders AlphaGo and man of the year in the science journals Science and Nature, confessed that Master(P) is indeed not a man, and artificial intelligence.

A professional go player, Niall Cardin notes that the work of artificial intelligence useful for the gaming community:
It’s wonderful that people have played a series of games with AlphaGo that it is now possible to carefully study and browse. They have a lot of very unexpected and fresh moves, which are extremely interesting for us.

Hassabis, said that soon will come a new version AlphaGo. In the summer it will compete with the strongest players in the world.