The ancient board game Go has generated tremendous interest in artificial intelligence, at least for people who are trying to develop capable computer Go programs. A recent tournament for such programs was held in Edmonton, Alberta, and showed the relative complexity in creating for Go the type of strong game-playing program that beat chess champion Gary Kasparov five years ago. Whereas chess requires more pure calculation, Go involves intuition and pattern-matching, very difficult tasks for a computer.
Dr. Michael Reiss, a computer scientist who wrote a previously winning program, says the challenge for a computer in recognizing patterns in the game of Go is the same posed when trying to distinguish the shape of a bicycle from that of a chair. That is, Go programs must be able to make decisions that people naturally make without much effort. Because of this, experts who write Go programs must necessarily be good Go players, and be able to decode their own complex thinking processes.
"I think
in the
long run the only way to write a strong Go program is to have it
learn from
its own mistakes, which is classic AI, and no one knows how to do
that
yet," says David Fotland, whose Go program won the Edmonton
tournament.
(http://www.nytimes.com/2002/08/01/technology/circuits/01GONE.html
Access to this site is free; however, first-time visitors must
register.)