EUGENE–A bridge program coming out of the University of Oregon’s Computational Intelligence Research Laboratory is aiming, with the help of artificial intelligence, to do for bridge what IBM’s Deep Blue did to chess in general and to grandmaster Garry Kasparov in particular.

IBM’s chess playing computer soundly defeated Kasparov, the best human chess player in the world. Oregon artificial intelligence researcher Matthew Ginsberg developed the bridge software program called GIB and nicknamed it "Deep Green" after one of the UO Ducks’ team colors. GIB has dominated the computer bridge scene since its introduction in 1998.

Ginsberg, a fellow of the American Association for Artificial Intelligence, is pleased that A.I. has been so central in his work on GIB.

"A.I. is a performance discipline, and GIB is a performance program," he says. "Many of the techniques GIB uses started out as theoretical ideas in our Oregon lab, and it’s always nice when cutting-edge research gets deployed in a commercial program."

GIB finished 12th among human competitors in one phase of the 1998 world bridge championships. Now Ginsberg and longtime colleague Fred Gitelman are setting their sights squarely on human targets in a series of matches co-located with the North American bridge championships over the next few years.

The matches will pit the Deep Green team of two GIBs, paired with Gitelman and his regular bridge partner Brad Moss, against teams of four to six human players who will need to win increasingly prestigious bridge events to earn the right to challenge the half-human, half-machine cyborg team.

These contests apparently will be GIB’s only opportunity to play against humans in formal competition. The American Contract Bridge League (ACBL) flatly turned down Ginsberg’s request for permission to enter GIB in the organization’s events.

"Our players come to a tournament expecting to compete against other people," said Gary Blaiss, chief tournament director for the ACBL. "I do not believe a mechanical/electronic horse would be permitted entry to the Kentucky Derby."

Ginsberg, meanwhile, doesn’t view this as strictly a GIB vs. human contest, citing the presence of Moss and Gitelman as Deep Green teammates. Computers and software are certainly no stranger to Gitelman, a frequent bridge partner of Microsoft’s Bill Gates and owner of a bridge software company himself.

"Machines should be our collaborators, not our competitors," says Gitelman. "Our strengths are complementary. That’s why we decided against setting this up as a series of man versus machine matches; it’s more man and machine together versus man on his own."

Ginsberg echoes these sentiments.

"The most striking thing about GIB or Deep Blue," he says, "is how these computer programs, using artificial intelligence, play their games so differently than we do. Whatever the movies such as ‘A.I.’ would have you believe, artificial intelligence isn’t about building artificial people. It’s about building systems that perform well, typically in quite alien ways."

Alien or not, it’s GIB’s performance that will matter most when the first match gets underway in Toronto at the end of July. Beating the machines and humans team will net the human team a $10,000 prize. Losing allows them to join Kasparov in the record books.

—30—

#P-7634/Local,OrSci,Top25Sci,PDX