Two moves nobody expected

More than 200 million people watched the five-game match. For many viewers, it became a test of something larger than one game.

In Game 2, AlphaGo placed its thirty-seventh move on the fifth line of the board. Professional commentators first struggled to explain it. AlphaGo's own model assigned only a 1 in 10,000 chance that a human would choose that move.

The move was not random. Much later, its position helped AlphaGo win. Lee Sedol called it creative. The system had found a strong move outside the patterns that centuries of recorded human play had made familiar.

The pressure on Lee was immense. The documentary records one observer saying that he was no longer playing only for himself or his country. People had made him a representative of the human race. After losing the first three games, he returned for Game 4.

Lee's seventy-eighth move was also one AlphaGo estimated at roughly 1 in 10,000. But it did not force a win. Professional analysis found a reply that could have kept AlphaGo ahead.

AlphaGo did not find that reply during the game. It made its mistake at move 79, while its still estimated about a 70 percent chance of winning. It did not register the trouble until around move 87. Then its estimate fell sharply, and it played a series of weak moves. Lee held the advantage and won.

Move 78 mattered because of what it demanded next. Lee found a move that AlphaGo had barely considered, and the system's judgement unravelled after it.

Players studied both moves after the match. The board had not changed. Their map of what could work on it had.

Move 37 did not prove that a machine creates as a person does. It revealed a strong possibility that human tradition had overlooked. Move 78 revealed something different: a weakness hidden by AlphaGo's otherwise overwhelming play.