计算机处理围棋复杂的能力压倒了人类

Computer overwhelmed human in handling the complexity of the game of Go

2016年3月以及2017年5月,在与李世石和柯洁两位最顶尖人类棋手的两次围棋人机大战中,Alpha Go分别以4:1和3:0的比分获胜。围棋界从对计算机围棋评价不高,到承认计算机已经彻底战胜了人类棋手,只有短短的一年多时间。文章介绍蒙特卡罗树形搜索、策略网络、价值网络、强化学习等围棋算法思想,回顾计算机围棋算法不断发展直至处理复杂的能力超过人类棋手的历程,并展望人工智能对围棋与社会的影响。

In March 2016 and May 2017, the computer AlphaGo defeated two top profes- sionals Lec Sedol and Ke Jie by 4：1 and 3：0, respectively, in the game of Go. In 2015 the Go community still cstimated the computer＇s ability to be fairly limited, but now admit that humans are completely defeated. We shall introduce important algorithm concepts such as Monte-Carlo tree search, policy network, value network, reinforcement learning, etc., review the breakthroughs of the computer Go algorithm that led to its final victory over humans, and discuss the impact of artificial intelligence on Go and society in general.