As was previously announced, Google’s DeepMind project, AlphaGo took on the world’s best Go player, Lee Sedol in the first of five matches yesterday. And it won. While its first victory against a different player was important, beating the 18-time world champion is another matter entirely. This is a momentous victory for AlphaGo, and for the machine learning industry as a whole.
Go is notoriously difficult to play from a machine’s perspective. There are so many possible moves, and the creativity of play so perplexing, teaching a machine to play against a human is incredibly difficult. But with Google’s DeepMind project, engineers set about building a machine which you could be taught to learn. It wasn’t simply a case of teaching it all the moves, but rather teaching it to think for itself and learn how to counter moves by the opponent.
AlphaGo learned by studying games played by humans, analyzing 30 million moves made by players before beginning to play its own internal games. In many ways, it was these internal games that were the most important in learning how to beat an unpredictable human player. It invented its own strategies, and was able to beat a person 5-0 when it last played a professional Go player.
Upon losing the first game yesterday, 33 year-old Lee stated, “I was very surprised because I did not think that I would lose the game. A mistake I made at the very beginning lasted until the very last.” He also described the machine’s strategy as “excellent” from the first play. As reported by The Guardian:
Yoo Chang-hyuk, another South Korean Go master who commentated on the game, described the result as a big shock said that Lee appeared to have been shaken at one point. Hundreds of thousands of people watched the game live on TV and YouTube.
AlphaGo still has four games remaining against Lee Sedol, which will end on Tuesday. While many will be surprised that a machine was able to beat the best player in the world, I think the question now is: Can Lee Sedol win at least one match against an AI-powered machine which doesn’t seem capable of losing.
Image Credit: Ahn Young-joon/AP
FTC: We use income earning auto affiliate links. More.