What made AlphaGo's move 37 so famous?
In 2016 a computer played a Go move most professionals would never have considered, and the world's best player needed a long think to answer it.
▶ Start the storyMove 37 was the moment a machine seemed to show creativity. In March 2016, in the second game of its match against Lee Sedol in Seoul, DeepMind's AlphaGo placed a stone that the commentator Michael Redmond called "creative" and "unique": a move most professional players would not even have considered. Lee, one of the strongest players in Go history, took an unusually long time to reply. AlphaGo won that game, and the match 4–1.
Go had long been seen as a far harder challenge for computers than chess, a game said to depend on intuition. Before 2015 the best programs played at amateur level, and some researchers claimed computers would never beat top humans. AlphaGo learned differently from older programs. It first studied around 30 million moves from strong human games, then improved by playing huge numbers of games against copies of itself. Its creators admitted they had no idea what moves it would come up with.
That is why its play looked strange. AlphaGo does not try to win by a big margin; it tries to maximize its chance of winning at all. Professionals said its odd moves looked like mistakes at first sight and like deliberate strategy in hindsight.
Humanity had its answer too. In game 4, Lee played his own unforeseen move 78, hailed as a "divine move", and AlphaGo's play fell apart. It was his only win. In 2019 Lee retired, calling AI "an entity that cannot be defeated".
4–1
Quiz me
0/3
Recap
AlphaGo learned from human games and then from playing itself, aiming to maximize its chance of winning rather than its margin.
Surprising fact · Lee's only win came from his own unforeseen move 78, nicknamed the divine move.
Sources (3)
No source, no claim. Every fact in this lesson (20 claims) cites at least one of these.