The Match That Changed Everything
In March 2016, DeepMind's AlphaGo faced Lee Sedol — one of the greatest Go players of his generation — in a five-game match that captured international attention far beyond the usual Go audience. The AI had already beaten European champion Fan Hui, but Lee Sedol was considered by many to be a different order of challenge entirely.
AlphaGo won games one, two, and three in succession. For the first time in history, the world's best human Go player appeared outmatched by a machine. Then came Game Four.
The Setup: A Match Under Pressure
Entering Game Four, Lee Sedol had publicly expressed confidence before the match began, and had now lost three straight games to an AI. The weight on him was immense. When Game Four began, many observers assumed AlphaGo would complete a 4-0 sweep.
The game unfolded across the board in the complex, multi-front fashion typical of top-level Go. AlphaGo appeared to be playing solidly, its characteristic slightly inhuman efficiency visible throughout the middle game.
Move 78: The "Hand of God"
Deep into the middle game, at move 78, Lee Sedol played a stone in an unexpected position — a wedge into a section of the board where AlphaGo had constructed what seemed like a robust framework. The move was so unexpected, so precisely calibrated to the specific weaknesses in AlphaGo's position, that the AI's win probability calculations — which had held steadily above 70% for much of the game — began to collapse.
AlphaGo began making errors. Move after move, it played at a level that human commentators recognised as confused — playing as though it could not reconcile what had happened with its model of the game. Lee Sedol went on to win Game Four, the only human player ever to defeat AlphaGo in official competition.
What Made Move 78 So Special?
Go commentators and AI researchers later analysed why this particular move was so devastating to AlphaGo:
- It was a "blind spot" move — AlphaGo had not seen sufficient training examples of this type of tesuji (tactical move) in this specific context, making it genuinely unprepared.
- It changed the strategic character of the game — The move didn't just win local stones; it fundamentally altered which areas of the board were alive, dead, or contested.
- It was human creativity at its finest — Lee Sedol himself said he had thought deeply before playing it, and wasn't certain it was the right move when he placed the stone.
Lee Sedol's Reaction
After winning Game Four, Lee Sedol described his feelings with characteristic directness. He said the victory felt more meaningful to him than many of his tournament wins — precisely because it came under such pressure and against an opponent that had seemed invincible. He later reflected that the experience changed how he thought about Go itself: it revealed that even a superhuman AI could have specific, exploitable weaknesses.
The Broader Significance
Game Four of AlphaGo vs. Lee Sedol is now one of the most studied games in modern Go for several reasons:
- It demonstrated that human intuition and creativity could, under the right circumstances, find moves that defeat algorithmic play.
- It became a touchstone in the broader conversation about AI — evidence that "superhuman" does not mean "infallible."
- It gave the Go community a moment of profound pride and inspiration, even as AlphaGo won the overall match 4-1.
A Moment Preserved
The full game record (SGF) of Game Four is freely available online and is worth loading into any analysis tool. Replaying the game through move 78 and watching the subsequent collapse in AlphaGo's play is a remarkable experience — a window into both the extraordinary capabilities and the specific limitations of the AI systems of that era, and a celebration of what human Go players can achieve at their best.