A milestone in Artificial Intelligence and Go was reached recently with science journal ‘Nature’ publishing a paper detailing the first victory of a computer program over a professional Go player.
On October 9 of last year five unofficial games and five official games were played between European professional player Fan Hui and Google Deepmind’s AlphaGo program. The result of the ten matches was decisive. Fan Hui won two unofficial games against Google’s artificial intelligence system, however AlphaGo prevailed in every official match played.
Go has proven a difficult game for computers to succeed at, largely due to the great number of available options presented in any given move. Alpha Go, developed by the Google Deepmind team, has used a new method to gain mastery over the game though. AlphaGo uses ‘deep neural networks’, allowing the artificial intelligence program to learn from professional games and apply and fortify this learnt behaviour in subsequent games. It’s clear that the new approach to the game has been successful, with AlphaGo clocking up a 99.8% win rate against other Go programs, as well as beating Fan Hui, a 2-dan player who became a professional in 1996.
As a result of AlphaGo’s success, a match has been arranged against Lee Sedol, one of the world’s best professional Go players, and is set to occur in March. Most of the Go community is still backing Sedol, given the large skill gap that exists between a 2-dan player such as Fan Hui, and a 9-dan player such as Lee Sedol.
Despite this, after AlphaGo’s victory against Fan Hui, Sedol’s match is sure to gain the attention of Go players across the world.
And though the odds seem in favour of the human player, few can deny that the day when artificial intelligence programs outclass human players is closer than ever.
So what does all this complete computer mastery of Go mean to the average Go player? With any luck, it means new and exciting ways for the game to be played being revealed, with new and exciting challenges for us to solve appearing alongside.
A commentary and analysis of the AlphaGo games against Fan Hui are available on YouTube.