Deep Blue - Wikipedia

Deep Blue is IBM's chess playing computer which beat then-world champion Garry Kasparov in a match in 1997.

An earlier version of the system played Kasparov in February 1996, losing 4-2 (first win by Deep Blue was on February 10). The system was then heavily upgraded and played Kasparov again in May 1997, winning the six-game rematch 3.5-2.5, ending on May 11th.

The project was started as "Chiptest" at Carnegie Mellon University by Feng-hsiung Hsu; the computer system produced was named Deep Thought after the fictional computer of the same name. Hsu joined IBM in 1989 and worked with Murray Campbell on parallel processing problems. Deep Blue was developed out of this.

The system derives its playing strength mainly out of brute force computing power. It is a massively parallel, 32-node, RS/6000, SP-based computer system enhanced with 256 special purpose VLSI chess processors. Its chess playing program is written in C and runs under the AIX operating system. It is capable of evaluating 100,000,000 positions per second.

Its evaluation function was initially written in a generalized form, with many to-be-determined parameters (e.g.: how important is a safe king position compared to a space advantage in the center, etc.). The optimal values for these parameters were then determined by the system itself, by analyzing thousands of master games. Before the second match, the chess knowledge of the program was fine tuned by grandmaster Joel Benjamin. The opening library was provided by the grandmasters Miguel Illescas, John Fedorovich and Nick De Firmian.

After the lost match, Kasparov said that he sometimes saw deep intelligence and creativity in the machine's moves, which he could not understand. He also suggested that humans may have helped the machine during the match. He demanded a rematch, but IBM declined and retired Deep Blue.

In part these allegations were correct. The rules provided for the engineers to modify the program between games, an opportunity they took with abandon. The code was modified between games to understand Kasparov's playstyle better, allowing it to avoid a trap in the final game that the AI had fallen for twice before.

External Links

IBM Research pages on Deep Blue at http://www.research.ibm.com/deepblue/