Mastering Atari, Go, Chess and Shogi by Planning with a Learned Model

MuZero replaces the perfect simulator inside AlphaZero-style search with a learned latent model that predicts only reward, policy, and value. Follow the three learned functions, the search-to-training loop, the Atari evidence, and the precise boundary behind the phrase without rules.

Reading focus: Why a planning model can ignore future pixels and preserve only the quantities tree search consumes. How representation, dynamics, and prediction functions let Monte Carlo tree search expand an imagined latent tree. What the board-game and Atari results establish, and why deterministic dynamics, compute, and model error still bound the result.

Nature 2020. Julian Schrittwieser et al.. 13 min read, hard difficulty.