Remembering Deep Blue's Surprising Chess Triumph Over A World Champion

Illustration for article titled Remembering Deep Blue's Surprising Chess Triumph Over A World Champion

Nineteen years ago today, IBM's Deep Blue computer made history by defeating reigning world chess champ Garry Kasparov.


At that time, Kasparov had been world champion since 1985, a title he held until 2000 (he retired in 2005). Deep Blue's first iteration also dated to 1985. explains:

[In 1985], Carnegie Mellon University doctoral student Feng Hsiung Hsu began developing a chess-playing computer called "ChipTest." Hsu was joined on the project by Thomas Anantharaman and Murray Campbell and the computer later came to be known as "Deep Thought," after a machine in the science-fiction novel The Hitchhiker's Guide to the Galaxy. The men were later hired by IBM, where they continued to work on the chess-playing computer. In 1989, Gary Kasparov easily trounced Deep Thought when they met for a 2-game match. Developers continued to refine the supercomputer, which in 1993 was renamed "Deep Blue," a combination of Deep Thought and Big Blue, IBM's nickname.

Despite his initial defeat against Deep Blue, Kasparov did go on to win the six-game match in 1996 (final score: 4-2, human) ... pretty impressive, considering the computer's ability to discern "100 million different chess positions per second."

A rematch the next year didn't end so well for Kasparov; he was defeated by a new, improved Deep Blue in another six-game match, which proved controversial when Kasparov accused IBM of constructing the machine specifically to beat him.

The below video offers a detailed analysis of the first game in the 1996 series.

Photo via Flickr Creative Commons user JasonBrown2013.




At the time it looked like a mere publicity stunt by IBM. The program was being modified between games by human masters, and may in fact have been modified during the games, which would have been cheating. IBM never revealed the logs of what happened to the machine during the games, and dismantled the machine itself. Some have accused the company of doing it to pump up their stock value.

This kind of scam characterizes a lot of AI promotion. This machine wasn't even vaguely intelligent - it was using custom hardware to do brute-force move extrapolation. Nothing about this promoted actual machine learning. It was a modern version of the Mechanical Turk. It really makes me wonder about IBM's later stunt of pitting the data miner Watson against the Jeopardy champion.