Google AI beats Lee Se-dol in ancient board game

Sean Callaghan, Sports Reporter

Go is an ancient board game—invented, legend has it, by the mythical First Emperor of China. It is said that it was made for the instruction of his son. It is played all over East Asia, where it occupies roughly the same position as chess does in the West. It is popular with computer scientists, too. For AI researchers, the idea of cracking Go has become an obsession. Other games have fallen over the years—most notably when, in 1997, one of the best chess players in history, Garry Kasparov, lost to a machine called Deep Blue. Modern chess programs are better than any human.

The rules of Go are simple. The players are Black and White, each given a bowl of stones filled with their color. Black starts. Players take turns to place a stone on any unoccupied space of a 19×19 grid of vertical and horizontal lines. The aim is to use the stones to claim territory. In the version being played by Mr Lee and AlphaGo each stone, and each surrounded intersection, is a point towards the final score. Stones surrounded by enemy stones are captured and removed. If an infinite loop of capture and recapture, known as Ko, becomes possible, a player is not allowed to recapture immediately, but must first play elsewhere. Play carries on until neither player wants to continue.

AlphaGo is a Google piece of Artificial Intelligence, programmed to play Go. This past week, AlphaGo beat the world leader. Google’s DeepMind AlphaGo program beat South Korea’s Lee Se-dol in the first of a series of games in Seoul. Several months ago, AlphaGo beat the Euro Champion. Mr. Lee had not expected a loss, “I was very surprised because I did not think that I would lose the game,” said Mr Lee. He also said “A mistake I made at the very beginning lasted until the very last.”

Print Friendly, PDF & Email