AlphaGo: The Original Move 37
March 2016. A black stone is placed on a fifth-line coordinate. 2,500 years of human intuition is shattered in a single algorithmic decision.
The game of Go contains more possible board configurations than there are atoms in the observable universe (roughly 10170 states). For decades, it was considered the holy grail of artificial intelligence research—a domain where brute-force computation was impossible and human "intuition" reigned supreme. Then came AlphaGo.
In Game 2 of the historic match against 18-time world champion Lee Sedol, AlphaGo played Move 37. It placed a black stone on the right-hand side of the board (coordinate Q10). It was a shoulder hit on a fifth line, a play so unorthodox that commentators initially assumed the human operator had misclicked.
"I thought AlphaGo was based on probability calculations and that it was merely a machine. But when I saw this move, I changed my mind. Surely, AlphaGo is creative." — Lee Sedol
The Mechanics of Alien Intuition
AlphaGo did not calculate Move 37 through an exhaustive search of all possible futures. It evaluated the move using a combination of deep neural networks: a policy network to narrow the search space to a handful of promising moves, and a value network to estimate the probability of winning from those positions.
When evaluating Move 37, AlphaGo's internal metrics showed a profound divergence from human baselines. The policy network, trained on human expert games, estimated that a human player would make that specific move with a probability of exactly 1 in 10,000. Yet the value network, which evaluated the raw win probability, saw a path to victory.
| Metric | Human Baseline | AlphaGo Assessment |
|---|---|---|
| Probability of human playing Q10 | 0.01% | Irrelevant |
| Win probability delta (Turn 37 to 80) | Negative (assumed) | +12.4% |
| Global influence | Local loss | Center-board dominance |
From AlphaGo to Zero
The original AlphaGo was bootstrapped on human data—it learned by observing human masters. But its successor, AlphaGo Zero, discarded human knowledge entirely. Given only the rules of the game and playing millions of matches against itself, AlphaGo Zero surpassed the original system in just 40 days.
This is the crux of Creativity vs Computation. By removing human bias, the machine discovered entirely new opening theories and strategies that humans had never conceptualized.