- Google DeepMind developed a new AI that learned chess from Stockfish, a powerful chess engine, without using search or self-play.
- The AI analyzed 15 billion board states and the moves that Stockfish would make on these boards.
- The result is an AI that can play at the level of a human grandmaster without ever having played a full game.
- The AI is small and fast, capable of making 20 moves per second on a personal computer with a $200 graphics card.
- The goal of this work was not to create a strong chess engine, but to demonstrate that a transformer type neural network can learn the expertise of a master by just watching the master at work.
- This concept could be used to create self-driving cars, new ray tracing algorithms, and more.
via DeepMind’s New AI Saw 15,000,000,000 Chess Boards!