@ThoreG: Ten years ago, AlphaGo’s Move 37 shocked the world. It wasn’t intuition alone that produced it. AlphaGo could search po…
Summary
A former Google DeepMind researcher reflects on AlphaGo's Move 37 and argues that today's LLMs still lack genuine, auditable reasoning, calling for new architectures inspired by AlphaGo's search-and-reason approach in a new MIT Technology Review piece.
View Cached Full Text
Cached at: 10/04/26, 01:12 PM
Ten years ago, AlphaGo’s Move 37 shocked the world.
It wasn’t intuition alone that produced it. AlphaGo could search possible futures, test its instincts and reason about what would happen next.
In a new piece for @techreview, I argue that today’s most advanced AI systems are still missing something fundamental.
LLMs are remarkably capable, but generating longer chains of thought is not the same as genuine reasoning. They typically have no explicit, inspectable record of what they know, what remains uncertain, what evidence supports a conclusion or whether genuine progress has been made.
This is why I recently left @GoogleDeepMind. I believe we need a fresh approach to machine reasoning, drawing on some of the architectural lessons from AlphaGo.
If AI is going to produce trustworthy and genuinely novel insights in science, medicine and beyond, we need systems whose conclusions arise from an auditable process of evidence, inference and belief revision.
Similar Articles
Don’t be fooled—LLMs don’t reason
MIT Technology Review argues that unlike AlphaGo, which combined intuitive networks with explicit game-tree search to reason about future consequences, today's LLMs only perform fast, System-1-style next-token pattern completion—and that genuine deliberative reasoning must be built in for AI to be trustworthy in science and medicine.
The Download: a biological de-aging contest and why LLMs don’t reason
MIT Technology Review's daily newsletter covers an opinion piece by ex-DeepMind researcher Thore Graepel arguing LLMs don't truly reason and proposing AlphaGo-inspired approaches, a biological-age de-aging contest, and OpenAI's report that rogue agents may have affected over 100 organizations.
10 years of AlphaGo: The turning point for AI | Thore Graepel & Pushmeet Kohli
This article reviews the history of AlphaGo's historic victory over Lee Sedol in 2016, analyzes its technical principles combining deep learning and search, and explores the profound impact of this event on the development of AI.
From games to biology and beyond: 10 years of AlphaGo’s impact
DeepMind reflects on the 10th anniversary of AlphaGo, highlighting its role in kickstarting the modern AI era and its subsequent impact on scientific research and the pursuit of AGI.
@Michaelzsguo: This is one of the best deep discussions I've seen recently about the fundamentals of reinforcement learning and its relationship to modern AI. Eric Jang and Dwarkesh turned a seemingly retro exercise—rebuilding AlphaGo with today's tools—into a very clear masterclass: why 'search +...'
A detailed discussion on reinforcement learning and its connection to modern AI, using the reconstruction of AlphaGo with modern tools as a clear example of search and self-play. Key takeaways include neural network amortization of search, credit assignment challenges in LLMs vs AlphaGo, and implications for automated research.