AlphaGo
- Date
- 2016 · confidence high
- Place
- Google DeepMind, London, United Kingdom; match in Seoul, South Korea
- Credited to
- DeepMind (Demis Hassabis, David Silver); defeat of Lee Sedol 4-1 in March 2016; AlphaZero (2017) learned without human data
- Field
- Knowledge & symbols
What it is
System that combined deep neural networks, reinforcement learning and tree search to defeat the world go champion, a game considered beyond the reach of machines for decades. Move 37 showed non-human creativity.
What it led to
- Chinese state investment in AI after AlphaGo's «Sputnik moment»
- Reinforcement learning applied to science (AlphaFold) and to language models
- The use of AI to discover strategies unknown to humans
Built on
- Deep Blue defeats the world chess champion1997
- AlexNet and the deep learning revolution2012
- reinforcement learning (TD-Gammon)
- Monte Carlo tree search
Made possible
- AlphaFold protein structure prediction2020
- Proof assistants (Coq, Lean) and AI mathematics (AlphaProof)1989-2024
- Autonomous AI agents and reasoning models2023-2025
- AlphaZero
- reinforcement learning from human feedback