Thomas J. Watson Research Center, IBM Corp.Yorktown Heights, N.Y.

Doug Bartholomew, Samuel Greengard, Glenn Hasek, John Jesitus, Scott Leibs, Kristin Ohlson, Robert Patton, Barb Schmitz, Tim Stevens, and John Teresko contributed to this article. When an IBM RS/6000 SP parallel supercomputer defeated chess Grand Master Garry Kasparov last May, the triumph was not of machine over man, but rather, the latest victory in mankind's continuing struggle to create technology that adds to the quality of human life. The game was simply the proving ground for a hardware and software alliance that established a new performance benchmark for "thinking" machines. Because Deep Blue was able to evaluate up to 50 billion positions on a chess board in the three minutes allocated for each move in tournament play, it won, but that is also the kind of performance needed to win in such diverse areas as data mining, weather forecasting, financial risk assessment, and complex simulations such as molecular dynamics used in pharmaceutical research. What made Deep Blue's victory possible was the combination of factors created by a team with a diverse set of capabilities. Even the most powerful computers, of course, do not actually "think." What they do is follow instructions. The team combined the know-how of Chess Grand Master Joel Benjamin with the experience and expertise of a group of research scientists and engineers headed by C.J. Tan. The RS/6000 SP, while extremely powerful, was not developed to defeat Kasparov. Similar systems are at work at more than 2,800 sites on hundreds of different scientific and commercial tasks. What IBM did for the match involved specialized algorithms and custom chips developed over months of work as computer experts analyzed the results of games between Deep Blue and two grand masters. When it comes to projecting all possible moves far into the future, no human player, not even a Kasparov, can match a supercomputer. But the human player's abilities go far beyond mere number crunching. A player at Kasparov's level can rule out most moves through instinct, experience, or intuition. Call it what you will, human thought transcends mere digital computation. But the combination of superior number crunching, the translated experience of the chess grand masters on the Deep Blue team, and the fact that a machine cannot be rattled, confused, or otherwise thrown off its stride by defeat, gave the victory to Deep Blue. But for those who think that the victory is a defeat for humankind, consider these words from Tokyo-based Chuck Goto, managing director of Smith-Barney International in Tokyo, written shortly after Kasparov's loss: "Whether we view computers as a threat to humankind and our livelihood, or a divine gift to better our lives, depends on how well we know and use them." Following the match, Kasparov himself suggested that chess players in the future might use computers to assist their play. And why not? Few engineering students would now take an important exam without a pocket calculator. Not long ago that would have meant automatic failure. The possibilities demonstrated by Deep Blue do not threaten the highest levels of human accomplishment -- creating great art, composing great symphonies, or leading nations or enterprises. What computers can do is amplify the abilities of the human intellect with massive data crunching capacity. Data mining, for example, is a promising new field where supercomputers can deliver substantial advantages to the business enterprise. With data mining, significant, useful, and often unsuspected relationships can be discovered in massive amounts of data. According to Al Lill, research director at Gartner Group, data mining is one of the key advances that "will trigger an evolutionary leap, rather than an evolutionary crawl, in the power of information technology." Machines like the RS/6000 SP can process huge streams of data in parallel to identify relationships or test business hypotheses based on real information that has been accumulated by an enterprise during years of operation. High-speed parallel processing also can be harnessed to model trends in financial markets. IBM scientists believe that the SP's ability to rapidly carry out assessments can translate into market advantage for users.