Move 37
In plain wordsA surprising move played by the AI program AlphaGo against champion Lee Sedol in 2016. Experts first thought it a mistake; it proved brilliant, and the phrase now stands for a creative, unhuman idea from AI.
- 1
History
The 37th move of the second game of the March 2016 match between AlphaGoAlphaGo, prop. n.: A DeepMind program that in 2016 beat Lee Sedol, one of the world's best players of the board game Go, a decade or more before many experts expected a computer to manage it. and Lee Sedol, in which AlphaGo, playing black, placed a stone on the fifth line in a 'shoulder hit' that professional commentators at first took for an error, because convention favored the third or fourth line in that position. AlphaGo's own policy network had estimated that a human would choose the move with a probability of about one in ten thousand, yet it proved central to the program's victory. Lee took an unusually long time to reply. The move became the emblematic example of a machine finding a strategy outside human tradition rather than imitating it.
- 2
Figurativefigurative
By extension, any action by an AI system that experts find surprising, initially implausible and ultimately superior — taken as evidence of novel insight rather than recombination of training data. The phrase recurs in discussions of machine creativity, AI-driven scientific discovery and reinforcement learningreinforcement learning, n.: A way of teaching a computer program by trial and error: it tries actions, earns points for good results and penalties for bad ones, and gradually learns which choices pay off over time. from self-playself-play, n.: Training an AI by having it play against copies of itself, so it always has an opponent at its own level and can keep improving without human examples., and as a benchmark for whether models can produce genuinely new ideas.
We are still waiting for the Move 37 of mathematics — a proof idea no human would have tried.
Origin
The game was played in Seoul on 10 March 2016. The European champion Fan Hui, who had lost to AlphaGo months earlier, called the move beautiful and said no human would have played it; the 2017 documentary AlphaGo carried the episode to wider audiences.
Related
- AlphaGoAlphaGo, prop. n.: A DeepMind program that in 2016 beat Lee Sedol, one of the world's best players of the board game Go, a decade or more before many experts expected a computer to manage it.
- self-playself-play, n.: Training an AI by having it play against copies of itself, so it always has an opponent at its own level and can keep improving without human examples.
- reinforcement learningreinforcement learning, n.: A way of teaching a computer program by trial and error: it tries actions, earns points for good results and penalties for bad ones, and gradually learns which choices pay off over time.
- AlphaZeroAlphaZero, prop. n.: A 2017 DeepMind program that taught itself chess, shogi and Go from nothing but the rules by playing against itself, and within hours of training beat the strongest existing programs.
- AlphaProofAlphaProof, prop. n.: A Google DeepMind system that learned to write computer-checked mathematical proofs by trial and error. In 2024 it helped an AI reach silver-medal level at the International Mathematical Olympiad for the first time.
Mentioned in
AlphaGoAlphaGo, prop. n.: A DeepMind program that in 2016 beat Lee Sedol, one of the world's best players of the board game Go, a decade or more before many experts expected a computer to manage it., AlphaGo ZeroAlphaGo Zero, prop. n.: A program from DeepMind that taught itself to play the board game Go better than any human in 2017, starting from nothing but the rules and learning only by playing millions of games against itself.