Like many his age at the time, Keh-hsun Chen began playing the ancient Asian board game of Go when he was just an elementary school boy in Taipei, Taiwan.

As Chen grew older and advanced through school, his attention turned to more adult interests, such as computing. His thesis was on recursive functions and Turing machines, and he wrote a dissertation on degrees of unsolvability. He would eventually go on to become a professor of computer science.

Now at the University of North Carolina at Charlotte, where he goes by the name Ken and serves as director of graduate studies in the College of Information Technology, Chen has not forgotten his childhood days. He has developed a computer program of the 4,000-year-old game called Go Intellect.

Chen’s program matched artificial intelligence (AI) with computers from throughout the world in at the 8th Computer Olympiad held last fall in Graz, Austria. Go Intellect was the only program to win a medal in both the 19×19 and 9×9 categories. His previous versions won gold medals in 1990 in both categories, as well as a silver and bronze in 2002 for the 19×19 and 9×9, respectively.

Chen says he first found out about the Olympiad was from the International Computer Games Association (ICGA) Journal, for which he serves as the computer Go section editor. But the competition itself was not his only reason for attending.

“It was a good opportunity for computer Go researchers and programmers to exchange ideas and test each other’s progresses,” Chen tells Local Tech Wire. “Also, I promised the ICGA Journal to write an Olympiad 19×19 Go tournament report. Plus Graz, Austria, was the culture capital of Europe for the year 2003 … a nice beautiful town to visit–. I had many reasons to enter the Olympiad.”

Simple rules, complex results

Go has 30 million registered players worldwide … most in Asia … but is gaining popularity in the United States, where Chen has the highest amateur ranking.

The game is played on a 19×19 grid with one player using white and the other using black stones. The two players take turns placing their stones one at a time on empty board of intersecting points. Once placed, stones cannot be moved; but they can be captured if surrounded by an opponent’s stones. The objective is to secure more grid points than the opponent does.

The basic rules of the game are straightforward, but the possible outcomes can be wildly complicated, which according to Chen, makes Go the most difficult board game to program, as opposed to a game like chess. He says even after three decades of research, the strongest Go computer programs can only play at an intermediate level.

“Go is the hardest popular board game to program, because the positional evaluation is usually extremely hard for the machine,” explains Chen. “Plus, it has very high branching factor…over 200 on average…making the computer chess-type, full-board search approach powerless. Go is an excellent test bed for new techniques and approaches in artificial intelligence.”

Catching a ‘break’

AI has always been Chen’s main research interest and that’s where he intends to leave it for now. He is still in the first year of a three-year term in his current role at UNC-Charlotte, where he oversees more than 50 Ph.D. students and some 150 MS students, as well as teaching a graduate algorithm class.

“I don’t have any plan right now to start a company,” Chen says. “Busy enough.”

While his long-term goals for the future remain to make a “breakthrough” in AI and computer Go, Chen is already thinking about the 9th Computer Olympiad, set for July 4-12, 2004, in Ramat-Gan, Israel.

“I hope my program will win two Gold medals,” he says.


8th Computer Olympiad: