Computer Systems Laboratory Colloquium

11:00AM, Wednesday, July 8, 1998
Terman 156

Computer Chess: the Deep Blue Saga

Feng-hsiung Hsu
IBM T. J. Watson Research Center
fhh@watson.ibm.com

In 1972, when Bobby Fischer played Boris Spassky for the World Chess Championship, chess became front page news worldwide for the first time. The two matches in 1996 and 1997 between Deep Blue, an IBM RS/6000 supercomputer, and Gary Kasparov, the human World Chess Champion, were more than just front page news. The matches went beyond chess and computer chess. A collective human nerve was touched, and it resonated. The events redefined our view of ourselves and our relationship with the technology that we create. The success of Deep Blue was in part the success of technological advances, but more importantly, it was the result of the team work from a dedicated team of computer scientists. A team of tool makers. We human beings had built a new tool that gave us a way to best the greatest human chess mind in the world.

How did this new tool come about? The convergence of parallel supercomputing technology and the ever improving semiconductor technology, which enabled the Deep Blue chess accelerator chips, played an important role. But the chain of events triggered when Claude Shannon gave his computer chess lecture in 1949 planted the seed. By the time the precusor project to Deep Blue was started in 1985, the following fact had been known for several years. That is, a chess program will play much stronger when it runs faster. For the next ten years or so, a continuous drive to attain greater search speed was launched until in 1996, the first Deep Blue was put together and played its first match with Kasparov. The result of the first match proved that the basic technology was in place but something else was needed to defeat the human World Champion than just technology--better tuned domain knowledge and a better prepared team. The team learned from the first match and addressed the problems in the year that followed. When 1997 came, the newer Deep Blue was beating Grandmasters and the older version of Deep Blue consistently. Finally, in May 1997, 48 years after Shannon's lecture, Kasparov, the human World Champion, lost the rematch to Deep Blue and thus ended the quest for one of the oldest holy grails in computer science.

Where do we go from here? There is certainly more work that can be done in computer chess, but there is also more in life than just chess and computer chess. Deep Blue achieved its great speed through the combination of a general purpose parallel supercompuer with special purpose accelerator chips. The same approach could be applied to a number of fields, including molecular dynamics, datamining, and financial modeling. Alternatively, one could explore the possibility of special purpose accelerators for wider classes of applications. The emerging field of reconfigurable computing holds some promise for this line of attack. Last but not least, as the Deep Blue events demonstrated, there is no real limit to what we humans can achieve when we put our minds together. The future is full of possibilities. Maybe it is time for something completely new...

About the speaker:

Born in Keelung, Taiwan. Hsu got the nickname "Crazy Bird" in high school for his tendency to do "insane" stuff. The name has stuck since. While studying for his Ph.D. in computer science at Carnegie Mellon University, he started alone in 1985 a computer chess project that would later be known as Deep Thought. Deep Thought became the first computer to play at Grandmaster level in 1988, and the Deep Thought team received the Fredkin Intermediate Prize as a result. For his work on Deep Thought, Hsu received the ACM Grace Murray Hopper Award in 1990. Hsu joined IBM in 1989 to continue the computer chess work, acting as the system architect and principal designer of an unnamed new chess machine. In 1996, the new machine, named Deep Blue, lost its first match but won the first ever regulation game by a computer over the World Chess Champion. Deep Blue won the rematch in 1997, and Hsu, along with Murray Campbell and Joseph Hoane, was awarded the Fredkin Prize for building the first computer to defeat the World Chess Champion in a regulation match.

Hsu received his Ph.D. in Computer Science from Carnegie Mellon University in 1989 and his Bachelor degree in Electrical Engineering from National Taiwan University in 1980.

Contact information:

Feng-hsiung Hsu
IBM T. J. Watson Research Center, MS: 27-233
P.O. Box 218
Yorktown Heights, NY 10598
Phone: (914) 945-2067
Fax: (914) 945-4077
Email: fhh@watson.ibm.com