AlphaGo
AlphaGo is a series of ai programs developed by deepmind to play the board game Go. It was the first computer program to defeat a professional human Go player without handicaps, marking a significant milestone in Artificial Intelligence and machine-learning.
Key Milestones
- 2015: Defeated European Go champion Fan Hui.
- 2016: Defeated Lee Sedol, a nine-dan professional, in a five-game match.
- 2017: AlphaGo Zero achieved superhuman performance by training solely through self-play, surpassing previous versions.
Technical Architecture
- Utilizes deep-neural-networks combined with Monte Carlo Tree Search (MCTS).
- Employs Policy Networks to select moves and Value Networks to evaluate board positions.
- Transitioned from supervised learning on human games to machine-learning via self-play.
Related Research & Context
- While AlphaGo focused on game-playing AI, subsequent DeepMind research expanded into other generative and modeling domains. For instance, recent insights from Sander Dieleman discuss the architectural shifts toward large-scale generative models:
- See Dieleman’s DeepMind Insights: Building Large-Scale Diffusion Models for Image and Video for a comparison of technical approaches in generative AI.