sourced story
9-15 March 2016Primary source · 2 sourcesWell documented

AlphaGo defeats Lee Sedol at Go

A game with more positions than atoms in the universe falls to a machine a decade early

On the timeline · around 9-15 March 2016 · Recognition and ReasoningRecognition and ReasoningAlphaGo defeats Lee Sedol at Go20252026

Quick facts

Opponent
Lee Sedol, 18-time world Go champion
Result
AlphaGo won 4-1
Location
Seoul, South Korea

What happened

DeepMind's AlphaGo, developed under David Silver, played a five-game match against Lee Sedol, winner of 18 world Go titles and widely regarded as one of the greatest players of his generation, in Seoul, South Korea. AlphaGo combined a 'policy network,' trained first on human expert games and then refined through self-play, to propose promising moves, with a 'value network' to judge how good a board position was, both guiding a Monte Carlo tree search rather than exhaustively searching Go's roughly 10 to the 170th power possible positions. AlphaGo won the match 4 games to 1, watched by an estimated 200 million people worldwide. In game two, AlphaGo played a move, later called 'Move 37,' that professional commentators initially thought was a mistake because it violated centuries of accumulated Go strategy, and Lee Sedol needed nearly fifteen minutes to respond to it.

Why it matters

Go had long been considered a much harder game for computers than chess because of its vastly larger search space and reliance on pattern intuition rather than calculation, so AlphaGo's win, which experts had predicted was roughly a decade away, was read as evidence that deep learning combined with search could master intuition-heavy domains previously assumed to require distinctly human judgment.

How we know

DeepMind's own account of the match documents the final 4-1 score and Move 37 directly; Nature's news coverage of the match independently corroborates the same result and format.

Sources

See something wrong? . Corrections with a source get fixed fastest.

Part of a timelineArtificial Intelligence30 events · From a wartime theory of neurons to machines that write, paint, and fold proteinsView all →