(Yicai Global) Dec. 12 -- DeepMind Technologies Ltd. released AlphaGo Teach on its website last night and aims to inspire Go players to try new strategies and uncover ideas in the millennia-old game with its learning tool.
By unveiling the system, the Alphabet Inc. [NASDAQ:GOOGL] unit has honored the promise it made after AlphaGo won a three-game match against Chinese grandmaster Ke Jie to share its artificial intelligence program’s data with Go lovers.
“The AlphaGo teaching tool has more than 20,000 different joseki and over 370,000 moves,” Fan Hui, a European Go champion and the AI program’s first teacher, wrote on Weibo. “[It] analyzes and suggests different opening sequences from AlphaGo’s perspective, and offers winning percentage estimates for every move. Hopefully we’ll be able to learn from AlphaGo’s unique understanding of the game.”
“[I’ll] learn the game from scratch,” Ke commented with his repost of Fan’s message.
The tool is based on AlphaGo Master and is available in several languages including simplified Chinese, Fan said.
“This tool provides analysis of 6,000 of the most popular opening sequences from the recent history of Go, using data from 231,000 human games and 75 games AlphaGo played against human players,” DeepMind said.
“The teaching tool may not be the ‘standard answer,’ but it offers us a vast space for exploration,” said Chang Hao, a world champion and nine-dan player.
“We’d better forget those fixed sequences, and I feel that I’m about to take off again after seeing some of the advanced moves in the tool,” said Zhou Ruiyang, a nine-dan player who participated in the Future of Go Summit in Wuzhen in May.