Is this the sign Sarah Connor was trying to warn us about? The robot uprising could be just around the corner – Google’s DeepMind AI just beat the world’s best Go player.
Go is a board game mainly played in China, Japan and Korea. It’s been around for thousands of years, but because it has a ridiculous number of possible moves (loads more than chess) it’s been difficult for computers to master.
AlphaGo, the game-playing program created for Google’s DeepMind AI, squared off against Lee Sed-Ol last night, and came out victorious. After three and a half hours of taking turns to put black and white tiles on the playing grid in an effort to occupy more of the board, Sed-Ol threw in the towel with 28 minutes left on his game clock.
We’ll admit to not having a clue how to play Go, but according to the experts Sed-Ol played an aggressive opener and AlphaGo countered every attack perfectly.
At its peak there were over 100,000 people watching the live stream on YouTube – you can watch the whole thing if you missed it with the stream archive below.
AlphaGo beat the European Go champion last year, but Sed-Ol is ranked first in the world. The win is a massive breakthrough for AI research, because of how DeepMind has been programmed.
To put it another way, IBM’s Deep Blue beat Gary Kasparov at chess, but in noughts and crosses, it would get thrashed by a toddler. DeepMind is a more general kind of artificial intelligence, one that adapts and learns to master any task from scratch, just like humans and animals do.
Yep, it all sounds a lot like SkyNet to us as well.
It’s not over yet, though. This was just the first in a five game series. The second game will kick off later today, with the overall winner pocketing a cool US$1 million prize.
DeepMind is going to donate its winnings to charity should it come out on top, and after a decisive first victory, the team must be feeling confident.
RELATED › Next Big Thing: AR Superhelpers