Human champion certain he'll beat AI at ancient Chinese game

AP News
|
Posted: Feb 22, 2016 8:07 AM
Human champion certain he'll beat AI at ancient Chinese game

SEOUL, South Korea (AP) — The world champion of the ancient board game Go, South Korean Lee Sedol, expects to prevail in a match with Google's computer program AlphaGo next month, but he's not so sure he will be able to do so a year later.

For now, Lee is predicting a 5-0 or 4-1 victory in his favor.

AlphaGo defeated a professional Go player for the first time in October, something that experts had predicted would take a decade. The match, described in a paper released in the journal Nature last month, marked a significant advance in the development of artificial intelligence.

Lee, 32, said AlphaGo's October match showed the program's ability was still slightly lower than his. It has not had enough time to improve its skills.

"But if artificial intelligence continues to advance, in a year or two years, it will be really hard to guess the results," Lee said.

Computers have long surpassed humans in other games, including chess. But Go, known as weiqi in China and baduk in Korea, is considered the most challenging for artificial intelligence to master because of its intuitive nature and complexity.

Before AlphaGo, the Go community thought it would take a few generations for computers to match human players. So the October shut out of the European champion was "truly shocking," said Park Chi-moon, vice president of the Korean Baduk Association.

The winner of the five-game match starting March 9 in Seoul will receive a $1 million prize. If AlphaGo wins, the prize will be donated to charities, including UNICEF.

Demis Hassabis, CEO of AlphaGo developer Google DeepMind, said the program will enable smartphones to provide smarter help for people in the near future. Eventually, it will enable computers to help scientists solve some of the toughest real-world problems, such as disease analysis and climate modeling, he said.

Go originated in China more than 2,500 years ago. It involves two players who take turns putting markers on a checkerboard-like grid. The object is to take over more area on the board with the markers than one's opponent, and to capture the opponent's pieces by surrounding them.

___

This story has been changed to reflect a translator's correction that Lee sees AlphaGo's skill as slightly lower, not a few levels lower, than his.

___

Follow Youkyung Lee on Twitter at twitter.com/YKLeeAP