Google AI bests the world’s number-one human Go player

Avatar By Aaron Sims | 1 year ago

Twenty years after a supercomputer won a chess game against a human world champion, a Google-owned artificial-intelligence system has outdone the world’s top human in the board game Go. The system, DeepMind AlphaGo, played Go world champion Ke Jie and won two games back-to-back out of a three-game matchup.

“I’m a little bit sad, it’s a bit of a regret because I think I played pretty well,” Jie told reporters afterward.

DeepMind AlphaGo is a product of Google subsidiary DeepMind Technologies Limited. The system was built exclusively to play humans at Go, a strategy game that consists of a 19-by-19 grid upon which players take turns placing stones and vie to take control of the most territory.

AlphaGo previously played a human at Go in March 2016 and won, but its latest victory over Jie, the world’s number-one Go player, sets a new precedent for the system’s performance.

Humans and computers have squared off in other high-profile board games over the years. In 1998, IBM computer Depp Blue won a chess game against Garry Kasparov, considered to be the world’s greatest chess player.

Go is a greater challenge for a computer, however. Or for humans, for that matter: Observers describe it as one of the world’s most complex game. AlphaGo prepped for the match by analyzing hundreds of Go games and playing matches against itself.

DeepMind said that as it further refines artificial-intelligence technology, it hopes to deploy it in medicine and science. The company is already working with Great Britain’s National Health Service to develop new software-based tools for diagnosis.