AI device beats professional Go player

[BaalChatzaf]

There story is here: http://www.nbcnews.com/tech/innovation/google-go-playing-computer-beats-pro-first-time-n505481

A word about Go. Go is an oriental board game that has been around for 3000 years. It is older than chess. It is also more complex. Go makes Chess look like something between Tic Tac To and Checkers. For a description of the game see https://en.wikipedia.org/wiki/Go_(game). It is remarkable how simple the rules of the game are. They can be learned easily by an 8 year old kid. But the strategy is deep and far more complex than Chess.

The AI system AlphaGo beat the Euorpean Go champion and a match has been arranged between the world's champion (he is Japanese) and AlphaGo. It will take place sometime in March.

The combinatorics of Go are so daunting that a straightforward brute force tree searching algorithm is out of the question. An AI learning machine studied millions of Go games to refine a strategy which works on an interplay between policy and positional value (strategy and tactics respectively). That means that the learning algorithm will have more applications eventually than just beating hapless Go players.

Ba'al Chatzaf