AlphaGo Played a Move Human Pros Said Was a Mistake, Then Won
Game 2 of the Lee Sedol match, March 2016: Move 37 looked like an error to every commentator and decided the game by Move 211.
DeepMind's AlphaGo played Lee Sedol, then ranked among the strongest Go players in the world, over five games at the Four Seasons in Seoul between March 9 and 15, 2016. The match was the first time an AI system had played a top-rated human at full board, no handicap, with the world watching by livestream. AlphaGo had warmed up the previous October by beating European champion Fan Hui 5–0 in a private match, but Fan Hui was rated several stones below the open top.
The match's signature moment came in Game 2, on the 37th move. AlphaGo played a shoulder hit on the right side of the board — a move that, by long human convention, you do not play in that position. Commentator Michael Redmond, a 9-dan professional, paused, said "that's a very strange move," and then reconsidered out loud. Lee, in the playing room, stood up and walked outside for fifteen minutes before responding. When the game ended around Move 211, AlphaGo had won, and Move 37 had been the pivot.
Lee Sedol won Game 4 with his own brilliancy — the so-called Move 78, a wedge play that exploited a blind spot in AlphaGo's evaluation network and that the commentary team called "the hand of God." Final score was 4–1. DeepMind donated the $1 million purse to charity; Lee received $170,000 for participation and his win. The Korea Baduk Association awarded AlphaGo an honorary 9-dan rank.
Make Recess yours.
Sign in to save the ones you loved, never see the same thing twice, and tell us what you want more of.