Move 37: The Move That Changed How We Think About Intelligence
Zusammenfassung
In Game 2 of the 2016 match between AlphaGo and Go world champion Lee Sedol, AlphaGo played Move 37 — a move on the fifth line of the board that no professional player would have considered. Go commentators called it a mistake. It took approximately a minute for human analysis to recognize it as profound. AlphaGo won that game and the match 4-1. Move 37 is now cited as the moment AI demonstrated not just competence but creativity — the ability to find solutions outside the space of solutions humans had explored. The one game Sedol won — Game 4, Move 78 — is equally famous: the only move that defeated AlphaGo in the entire match.
The Match Context
The 2016 match between AlphaGo and Lee Sedol was covered in DeepMind and AlphaGo. Lee Sedol was one of the strongest human Go players alive, with 18 world titles. The match was held in Seoul in March 2016; each game was streamed live with professional commentary.
Before the match, most Go professionals believed AlphaGo would lose. The game’s complexity — more positions than atoms in the universe — had made it resistant to computer analysis for decades. Go programs had lagged far behind human play compared to chess programs. Expert estimates placed the time to superhuman AI at 10 years; the actual gap was months.
Move 37 Itself
In Game 2, on the 37th move, AlphaGo placed a stone on the 5th row — a “shoulder hit” on a Chinese opening formation. Fan Hui, a European Go champion who was present as a commentator, said: “That’s not a human move.” Lee Sedol left the room for 15 minutes.
AlphaGo’s analysis estimated there was approximately a 1 in 10,000 chance that a human player would have made that move. The move was unusual not because it was random but because it reflected a long-range strategic plan that human players had not catalogued. AlphaGo had discovered, through self-play against itself in millions of games, patterns and strategies that humans had not explored.
Go professionals reviewing the game after the match concluded that Move 37 was not only not a mistake but was one of the finest moves played in a professional-level game that year.
Move 78: The Human Counterpunch
In Game 4, Lee Sedol played Move 78 — a stone placement in an empty corner that AlphaGo had not seen coming. AlphaGo’s evaluation of the position shifted dramatically; it began playing suboptimally and resigned. This is the only game in the match that Lee Sedol won.
Go players call Move 78 “God’s Touch.” Sedol described it as the greatest move of his career. The move is notable because it demonstrates that even a superhuman AI system can have blind spots — patterns it has not encountered in training that, when exploited, produce unexpected failures.
The combination of Move 37 (AI creativity beyond human imagination) and Move 78 (human creativity finding AI blind spots) made the 2016 match the most philosophically significant human-versus-machine contest since Kasparov-Deep Blue.