Deep Blue

Deep Blue's lineage traces to the 1980s at Carnegie Mellon University, where Feng-hsiung Hsu and Murray Campbell developed a chess machine called ChipTest, followed by Deep Thought. IBM recruited them in 1989 to build a machine that could challenge the world's best. The project, initially named Deep Thought, was renamed Deep Blue as a play on IBM's nickname, "Big Blue." The system combined custom VLSI chess chips with a parallel architecture based on IBM RS/6000 SP supercomputers. Each chip could evaluate millions of positions per second, and the entire system evaluated about 200 million positions per second. The hardware was specifically optimized for chess, using a brute-force search algorithm enhanced with heuristics and a large opening book.

In February 1996, Deep Blue faced Garry Kasparov in a six-game match in Philadelphia. Kasparov won 4–2, but Deep Blue made history by winning the first game—the first time a world champion lost a game to a computer under tournament conditions.¹ IBM then upgraded the machine, doubling its speed and refining its evaluation function. The rematch took place in May 1997 in New York. Deep Blue won the deciding game six, securing a 3.5–2.5 match victory. Kasparov later claimed that IBM had cheated, citing the machine's human-like play in game two, but IBM denied any unfair advantage. The match was widely covered by the media and sparked intense debate about the implications of machine intelligence.

Deep Blue's architecture was a hybrid of custom hardware and software. Each of its 30 IBM RS/6000 SP nodes contained 16 chess chips, totaling 480 chips. The chips implemented a hardware search that evaluated up to 200 million positions per second. The system used a minimax search with alpha-beta pruning, extended to a depth of 6 to 20 plies depending on the position. Its evaluation function considered material, piece placement, king safety, and pawn structure. After the 1997 match, Deep Blue was retired and disassembled. Its legacy includes influencing later chess engines like Stockfish and AlphaZero, though those use different approaches. Deep Blue is on display at the Smithsonian Institution and the Computer History Museum.²

Deep Blue's victory was not a pure brute-force triumph; it relied on a sophisticated opening book and an endgame database for positions with up to six pieces. The system also had a "personality" setting that allowed it to vary its play, which some observers found unsettling. A notable detail is that Deep Blue's evaluation function was hand-tuned by grandmaster Joel Benjamin, who was hired to help improve the machine's positional understanding. After the match, IBM refused Kasparov's request for a rematch, and the machine was never used again. The 1997 match was the subject of the 2003 documentary "Game Over: Kasparov and the Machine," which explored conspiracy theories. Deep Blue's success also inspired the creation of the "Deep Blue" chess program for personal computers, though it was unrelated to the original hardware.

Alpha-beta pruning

An optimization of the minimax algorithm that reduces the number of nodes evaluated in a search tree.

Minimax

A decision rule used in game theory to minimize the possible loss for a worst-case scenario.

Plies

A single move by one player in chess; a ply is half a move.

VLSI

Very-large-scale integration, the process of creating integrated circuits by combining thousands of transistors.