deepmind.md

DeepMind

• 返回上层目录
• AlphaGo-Zero: Mastering the Game of Go without Human Knowledge Nature2017
• AlphaStar: Grandmaster level in StarCraft II using multi-agent reinforcement learning Nature2019
• AlphaTensor: Discovering faster matrix multiplication algorithms with reinforcement learning Nature2022

===

DeepMind 研发的围棋 AI AlphaGo 是如何下棋的？

AlphaGo的第一作者David Silver还在MIT做post-doc的时候(也有可能是visit?)，曾经和我们组师兄合作利用机器学习和蒙特卡罗树搜索玩《文明2》。当时也有不小的轰动：

Human-level performance in 3D multiplayer games with population-based reinforcement learning

DeepMind 在多智能体强化学习方面又有了新进展，最新成果登上 Science 杂志！http://www.hzhcontrols.com/new-1161721.html

AlphaStar agent 24 JANUARY 2019

https://deepmind.google/discover/blog/alphastar-mastering-the-real-time-strategy-game-starcraft-ii/