After reaching the pinnacle of success this week, the Google AI AlphaGo is now retiring.
AlphaGo beat the world champion of Go, an ancient Chinese board game. The game is quite complex, more so than chess, even though its rules are quite simple. Players move black and white game pieces in an effort to cover more squares on the board than their opponent. Go's complexity is such that it has more possibilities than there are atoms in the universe.
19-year-old Ke Jie, who has been playing Go professionally since he was 10, is the Go world champion—or, at least he was, until AlphaGo beat him. Ke lost all matches against AlphaGo in a three-game series. Now, the Google AI has a game ranking higher than that of any human player. Experts apparently thought that AI won't be able to best human Go players for at least a decade still. It seems that their predictions were wrong.
When Ke lost the first match againt AlphaGo, he described the program as “godlike”. Because of the complexity of Go, people thought that players would need human intuition to win the game. It seems that this isn't exactly true.
Ke put up a valiant fight and only narrowly lost to the program. However, AlphaGo has proven itself to be a formidable opponent and has forced players to look at their own strategies in a different light.
Ke isn't the only high-ranking player that this Google AI has beaten. AlphaGo has also beaten Korean Lee Sedol, the “Roger Federer of Go”, four matches to one. Three-time Go champ Fan Hui, meanwhile, also lost to AlphaGo five to nil.
DeepMind, the Google subsidiary that developed AlphaGo, says that the AI has reached its peak, at least in Go competitions. After all, it has beaten the top Go players in the world. So what's next for this groundbreaking program?
The team behind AlphaGo is setting its sights on other challenges, now that it has surmounted the challenge of Go. AlphaGo is set to help newbies learn the highly complicated game and challenge veterans to think outside the box. AlphaGo's developers will also release a final research paper detailing the program's development over the past year.
According to a blog post by DeepMind CEO Demis Hassabis, the developers now aim to create “advanced general algorithms that could one day help scientists as they tackle some of our most complex problems, such as finding new cures for diseases, dramatically reducing energy consumption, or inventing revolutionary new materials”.
Go wasn't the Google AI's ultimate goal. Mastering the game was merely a stepping stone, and now, developers are moving forward.
Get weekly science updates in your inbox!