Computer has a go and beats pro player

I've just read over on Slashdot that a supercomputer has beaten a professional human player at the ancient boardgame, go, albeit with a 9-stone head start. It's a surprising result to those familiar with the game, since computers have so far proved no match for human players.

MoGo's performance stunned onlookers, including another go software programmer who said: "I'm shocked at the result. I really didn't expect the computer to win in a one-hour game."

Although it has similarities to chess in computational terms, go strategy is in practice much more complex. Its large board and few rules mean that a computer attempting to calculate a "tree" of possible future moves quickly creates an exponentially growing tangle. In the relatively short time available in a game, there isn't time to work out the best option.

Current AI techniques just aren't up to scratch. Increasing the speed with which calculations can be made is thought to be unlikely to lead to proper computer supremacy - although it seems to have played a major role in MoGo's victory. Instead many experts say novel ideas about how to give AIs some equivalent to the kind of intuition used by human players are needed. Easier said than done.

Tom Simonite, online technology editorLabels: artificial intelligence, computing, games

Posted by Tom at 5:15 PM