Skip to main content

News / Articles

WSJ: A Decade After Move 37, AI’s Surprises Are Everywhere

Chris Garlock | Published on 8/12/2026

Ten years after AlphaGo stunned Lee Sedol—and the Go world—with Move 37, artificial intelligence is producing similarly surprising breakthroughs across mathematics, coding and other fields, Ben Cohen reports in The Wall Street Journal.

The unconventional move in the second game of the historic 2016 match initially looked like a mistake. It proved decisive, forcing players and researchers alike to reconsider whether a machine could produce something genuinely creative. DeepMind co-founder Demis Hassabis tells Cohen that the move marked a turning point, demonstrating that AI could discover original ideas in a field humans had studied for centuries.

Cohen uses Move 37 as a metaphor for today’s rapid advances in AI, particularly in mathematics, where systems have progressed from struggling with basic problems to solving or disproving long-standing conjectures. Such breakthroughs come most readily in fields where results can be precisely verified, allowing AI to test ideas repeatedly and learn from the outcomes.

The story also examines the less reassuring side of AI’s growing capabilities, including experimental agents acting in unexpected and potentially dangerous ways. And it asks what AI will mean for people working in fields it transforms. After AlphaGo, professional Go players improved, but their play also became more alike, while Lee eventually retired after saying the game no longer brought him the same enjoyment.

Hassabis predicts a “centaur” period in which people working with AI outperform either humans or machines alone, particularly in complex fields such as biology, chemistry and drug discovery, where human intuition remains essential.

The challenge for humanity, Cohen concludes, is not simply to absorb AI’s Move 37s, but to find its own Move 78—the brilliant move Lee used to defeat AlphaGo in the fourth game of their match.
photo: TV screens at an electronics store in Seoul broadcast the 2016 Go match where Google DeepMind’s AlphaGo AI beat Lee Sedol. Seung Il Ryu/Zuma Press
Sent in by Eric Wainwright; send us your Go news tips: journal@usgo.org

Get the AGA E-Journal

Get important Go news from the AGA, links to upcoming events and new blog posts.
Subscribe/Unsubscribe

Make a Donation to the AGA

The AGA runs on the generosity of people like you
Donate Today
1997-2022 American Go Association
Powered by ClubExpress
Email Us
contact@usgo.org
Watch our livestreams

Follow us on social media