Google Beats Facebook to AI Board Game Breakthrough

Google artificial intelligence GO Facebook AI
The ancient Chinese game Go is considered one of the most complex board games in the world, with more possible board combinations than there are atoms in the universe. Nature

A Google artificial intelligence algorithm has beaten a professional human player at the ancient Chinese board game Go for the first time.

The search engine giant made the announcement of the AI breakthrough on Wednesday, just hours after Facebook revealed that six months of its own research into creating a Go-playing AI had failed to beat a human.

Go represents a much more significant challenge for artificial intelligence software than other board games like backgammon, chess and draughts due to the complexity of the game. In an average 150-move game, the number of possible board combinations is larger than the number of atoms in the universe.

Google’s London-based AI firm DeepMind developed a program called AlphaGo that mastered the game.

“I’ve always thought it would be a great challenge for computers to be able to play such an aesthetic game, such an intuitive game like Go,” said DeepMind founder Demis Hassabis. “A much greater challenge than it was to play chess.”

AlphaGo was able to beat European Go champion Fan Hui in five games out of five. In comparison, Facebook’s software was only ranked as “advanced amateur” and not “professional level,” according to a paper detailing the social network’s research efforts.

“Last year, the Facebook AI Research team started creating an AI that can learn to play Go,” Facebook CEO Mark Zuckerberg said in a post to the social network. “We’re getting close, and in the past six months we’ve built an AI that can make moves in as fast as 0.1 seconds and still be as good as previous systems.”

The paper revealed that Facebook’s AI bot sometimes played “pointless” moves and would need to be improved in order to beat professional human players.

Google now hopes to test AlphaGo against other top Go players from around the world, while also try to understand how its program could potentially be used to carry out new tasks.

“We’ve no idea how to do that. Not yet,” Hassabis said. “Most games are fun and were designed because they’re microcosms of some aspect of life and they may be slightly constrained or simplified in some way. But that makes them the perfect challenge for us as a stepping stone towards building general artificial intelligence.”