Ten years ago, a computer made a move that no human would have made.
That moment — the 37th move of a 2016 Go match between Google's AlphaGo program and a human champion — has become shorthand for the instant machines stopped merely imitating us and started surprising us. Writing in the Wall Street Journal, Ben Cohen revisits Move 37 as a watershed in artificial intelligence: a play that looked like a mistake to expert commentators and turned out to be brilliant.
Cohen's argument, as summarized by Techmeme, is that the same thing is now happening in mathematics. AI systems are producing genuinely unexpected results in a field where surprise is the whole point — where a single unfamiliar step can open a path nobody knew existed.
The Conversation, in a separate piece, puts it more bluntly: generative AI has changed mathematics forever, and the open question is where the discipline goes next.
Why Go and math make such a natural pair is worth spelling out. Both are closed systems with fixed rules and effectively infinite possibilities, which makes them ideal terrain for machines that can search further than any person can. Both also have a deep human culture of taste and intuition — a sense of which moves, or which proofs, are elegant. Move 37 was unsettling precisely because it violated that taste and was right anyway.
Why it matters: if AI can reliably find ideas that expert humans would never consider, mathematics becomes the first serious test of whether machines can be true collaborators in discovery rather than very fast calculators.