Reinforcement Learning · advanced · concept 192 of 311 · Updated Nov 4, 2026
AlphaZero and MuZero
AlphaZero (2017) learns chess, shogi and Go from the rules alone, by self-play plus Monte Carlo tree search guided by one network that outputs a move policy and a position value. MuZero (2019) drops the rules too: it learns a model of the game's dynamics in a latent space and plans with that, matching AlphaZero on board games and setting a new state of the art on 57 Atari games. MuZero was later applied to video compression at YouTube.
Top AI contributors on this
- Demis HassabisGoogle DeepMind (chair) / Isomorphic Labs
- Pieter AbbeelAmazon / UC Berkeley
- David SilverIneffable Intelligence / UCL
- John SchulmanThinking Machines Lab
- Sergey LevineUC Berkeley / Physical Intelligence
- Chelsea FinnStanford University / Physical Intelligence
Key terms
Where you meet it in the real world
Chess and Go engines, video compression (YouTube), chip design and scheduling research, planning in simulators, game AI
Videos
Google DeepMind · YouTube
Yannic Kilcher · YouTube
Google DeepMind · YouTube
Guides and articles
Wikipedia
Wikipedia
UCL / David Silver
Courses, papers, and more