Deep Blue
Definition
Deep Blue was IBM's chess computer that beat world champion Garry Kasparov in 1997. It was the first time a computer defeated a reigning world champion in a complex intellectual game under official tournament conditions.
Deep Blue's roots go back to the "ChipTest" project, which began in 1985 at Carnegie Mellon University, led by the researcher Feng-hsiung Hsu. The project moved to IBM, was renamed "Deep Thought," and finally, in 1989, got the name "Deep Blue." Deep Blue did not "learn" chess the way modern AI models learn. It relied on enormous computing power to examine millions of possible moves ahead (six to eight moves deep, and in some positions 20 or more), combined with detailed chess knowledge programmed by hand in advance by human chess experts. The 1997 version ran on an IBM RS/6000 SP supercomputer with 30 PowerPC 604e processors and 480 dedicated chess chips, and evaluated about 200 million chess positions per second. That performance reflected 11.38 GFLOPS (billions of floating-point operations per second) on the standard LINPACK benchmark.
In February 1996, Deep Blue played reigning world champion Garry Kasparov in Philadelphia and won the first game. It was the first time a computer beat a reigning world champion in a single game under regular tournament conditions. However, Kasparov recovered and won the match 4–2 (three wins, two draws and one loss to Deep Blue). The decisive meeting took place in May 1997 in New York. There, after major hardware upgrades, Deep Blue won the match 3.5–2.5. It was seen as a turning point in public awareness: in many people's eyes, the first proof that "computers can think" in a field considered the peak of human intelligence. It also established chess as a recognized and accepted benchmark for the abilities of artificial intelligence.
Despite the achievement, Deep Blue was built as a machine for chess only, with no general learning ability. This is unlike the approach of AlphaGo, about twenty years later, which succeeded at the far more complex game of Go by learning on its own from data and self-play, not just by brute-force search. Two parts of the machine are on display in museums today: one at the Smithsonian's National Museum of American History, and the other in the AI (artificial intelligence) and robotics exhibit of the Computer History Museum.
Found a mistake, or have a source we should add? Write to us