Sat, Dec 15, 2018 - Page 2 News List

Intelligent go software to boost nation’s performance

By Lin Chia-nan  /  Staff reporter

Taiwanese-Australian professional go player Joanne Missingham plays against the “CGI GO Intelligence” software program developed by National Chiao Tung University professor Wu I-chen’s lab during a lecture held by the Chinese Juvenile WeiChi Institute in Taipei on March 27, 2016.

Photo: Liao Chen-huei, Taipei Times

National Chiao Tung University professor Wu I-chen (吳毅成) presented an intelligent “Life-Learning System for Go” training and grading system for go players at the Future Tech Expo in Taipei yesterday.

The system is the result of a years-long endeavor he hopes could help players make faster progress.

The three-day expo, which ends today at Taipei World Trade Center Hall 3, showcases techniques related to artificial intelligence (AI) applications, biotechnology, medical devices, chemical engineering and materials science.

In the AI section, Wu introduced a smart learning system for go players that offers nearly 40 difficulty levels, from beginner to professional and above.

Many go players said they felt they were making no progress when playing against computer systems, which motivated him to develop his intelligent system, Wu said, adding that it could become a tool to assess players’ levels.

The system would not have to slow its computing when adjusting levels, as most other go systems do, and its steps are less predictable, said Wu, who runs a computer games and intelligence (CGI) laboratory at the university and started developing the system in 2015.

The lab last year received a NT$6.6 million (US$213,821) donation from Siliconware Precision Industries Co chairman and Haifong Weiqi Academy founder Bough Lin (林文伯), who signed a three-year contract with the university for a project to boost go players’ performance.

Weiqi (圍棋) is the Chinese word for go.

Cloud computing technology supplier Ubitus Inc provided Wu’s lab with graphic processing units.

A previous CGI go program designed by Wu’s team was ranked the world’s second-strongest in an international match in China last year, after it beat China’s Fine Art Go software and Japan’s DeepZenGo, but was later defeated by DeepZenGo, academy head coach Chou Chun-hsun (周俊勳) said.

Before the match, the Chinese and Japanese systems had been the second-strongest systems since Google’s AlphaGo retired, Chou added.

Considering that AlphaGo is able to train itself without studying human game records, Wu hopes that his system can also reduce reliance on human players, Wu said.

His lab would announce a new technique showing the system’s evolution next year, he added.

This story has been viewed 1771 times.

Comments will be moderated. Remarks containing abusive and obscene language, personal attacks of any kind or promotion will be removed and the user banned.

TOP top