September 18, 1997
Member of IBM team behind Deep Blue speaking at UW
IBM Deep Blue’s win over Garry Kasparov in May marked the first time a computer defeated a world chess champion in regulation play. On a deeper level, it stimulated a debate about the concept of thinking and opened a new era in technology. But Deep Blue isn’t done. IBM researchers are now focusing Deep Blue’s extraordinary processing power on other mind-boggling challenges.
Murray Campbell, a research scientist at the IBM T.J. Watson Research Center where Deep Blue was developed, will address the challenges and implications involved in the 50- year effort to put a computer atop the chess world in a free public lecture at the University of Washington. The talk is scheduled for 7:30 p.m. Sept. 22 in Kane Hall Room 120.
IBM Deep Blue is a research experiment in massive parallel computing designed to learn how to maximize the processing power of a system and harness it to tackle problems currently too difficult to solve with computers. Chess is used because it is a complex game with simple rules. Future applications of this technology include solving equally complex problems in pharmaceutical drug development, financial risk assessment and disease patterning.
The power behind Deep Blue is specialized hardware and software which resides on an IBM RS/6000 SP parallel supercomputer. It is capable of examining 200 million moves per second or 50 billion positions in the three minutes allocated for a single move in a chess game.
But the key to Deep Blue and the biggest challenge for its developers in preparing to face Kasparov was instilling human expertise, chess knowledge in this case, into the computer program, according to Campbell. “This is known as the knowledge bottleneck in artificial intelligence circles,” he explains. “My primary focus in Deep Blue in the last year was in improving the chess knowledge of the Deep Blue system, in part by working together with a chess Grandmaster.”
Another major challenge for supercomputers such as Deep Blue in playing chess and in tackling other complex problems is to coordinate many independent processors around a single problem, says Carl Ebeling, professor of computer science and engineering at the UW.
“The challenge of parallel computing for chess is coordinating the processors so that they’re not all doing the same work,” Ebeling explains. “Dr. Campbell is going to spend some time talking about artificial intelligence issues in general and to what extent chess applies to the larger problem of artificial intelligence.”
Campbell, a former chess champion of Alberta, Canada, has been involved with computer chess research for 15 years. He was awarded the Fredkin prize and the Allen Newell Research Excellence Medal for his work on IBM Deep Blue.
Campbell’s lecture is part of a regular seminar series on current topics in computer science sponsored by the UW Department of Computer Science & Engineering.
###
For more information, contact Marcy Holle of IBM Research Communications at (914) 945-3970 or Ebeling at (206) 543-9342 and ebeling@cs.washington.edu.
<!—at end of each paragraph insert
—>
