MMOOZZGG.com

Glossary

What was the Deep Blue chess computer?

In short

The Deep Blue chess computer was a machine built by IBM that won a match against world champion Garry Kasparov in May 1997.

History of AI · Deep Blue

MMOOZZGG.com editorsChecked: 1 min read

For a child

Imagine playing checkers with a friend who, before each move, tries a million options in their head. Your friend is not smarter than you, just very fast at counting. Long ago, engineers built such a huge machine for chess. It went through moves at an amazing speed and one day beat the world champion. People were surprised, even though the machine could not play any other game.

In plain words

It was a special computer that IBM built for one task: playing chess. It did not learn from experience like modern neural networks. The developers gave it rules for scoring a position, and the machine looked many moves ahead, about 200 million positions per second, and chose the move with the highest score. The machine lost its first match with Kasparov in 1996 and won the rematch a year later. It could do nothing except play chess.

For business

The story of Deep Blue shows that a narrow system can beat a human at one task long before general AI exists. For a company, this is a practical lesson: the gain comes not from a “smart machine in general” but from a precise task with clear rules and a measurable result. Many useful systems work this way today: demand forecasting, document checks, request routing. The second lesson: the machine was built on special hardware for the matches with the champion, and after the win it went to a museum. An expensive build for one demonstration rarely pays back. So before the start, count how many tasks the system will cover and what support will cost. A pilot tests these estimates.

For a specialist

IBM built Deep Blue; among the early developers, the company names Feng-hsiung Hsu and Murray Campbell. The system belonged to classical AI with no learning: game tree search by brute force and a position evaluation function written by people. For the rematch, IBM invited grandmasters to advise the team. According to IBM, the machine used 32 processors for coordinated parallel computing, evaluated 200 million positions per second, and performed 11.38 billion floating-point operations per second. In February 1996 in Philadelphia, Kasparov won the six-game match 4–2, although the machine took the first game. In May 1997 in New York, Deep Blue won 3.5–2.5, and the sixth game decided it. For games with far more positions, such as Go, brute-force search does not work: they needed neural networks and reinforcement learning.

Example

A chain of car service centers wants “AI for everything” and cannot choose a task. A consultant suggests starting the Deep Blue way: one narrow task with clear rules, such as picking a booking slot, and one metric, the share of requests booked without a manager.

Related terms

Questions and answers

When did Deep Blue beat Kasparov? The rematch took place in May 1997 in New York, and the computer won 3.5–2.5. A year earlier, in February 1996, Kasparov won the first match 4–2.
Was Deep Blue artificial intelligence? Yes, but in the old sense of the word. It did not learn from data.

It searched through moves by evaluation rules that people wrote, and it could only play chess. Modern systems such as language models learn from examples.

Sources

Not sure whether your company needs this?

AI solutions map
Contact us