Eric Jang – Building AlphaGo from scratch
AlphaGo developmentGo game rulesMonte Carlo Tree SearchDeep learningNeural networksReinforcement learningSelf-play trainingComputational complexityScaling lawsAutomated AI researchLLM trainingRobotics RLGame theoryOff-policy trainingOn-policy training
Eric Jang explains how to build AlphaGo from scratch using modern AI tools, detailing the game of Go's rules and the core Monte Carlo Tree Search (MCTS) algorithm. He describes how deep neural networks, specifically value and policy networks, enhance MCTS efficiency and enable AlphaGo's self-play training. The discussion also explores the broader implications of AlphaGo's success for understanding computational complexity, scaling laws in AI, and the challenges and opportunities in automating AI research.