Ancient history
Chess and computers go back a long way, perhaps because both appeal to rule-intensive, highly logical minds. In 1950, Claude Shannon, a scientist at Bell Laboratories, published the article that spawned the field of computer chess. (For a technical but complete history, see "Kasparov versus Deep Blue" in the bibliography.)

Shannon outlined how any chess computer would have to evaluate and choose future positions. He also gave some suggestions about how far into the future it would have to search -- a key consideration given the extremely limited talents of the early computers and the fiendish complexity of chess.

By that we mean this: In the middle of a game, when many pieces remain in play, each player typically has 30 or 40 moves. So after one move by each player (that's called two "plies," or one "move") the board could show about 1,000 positions. By another complete move, there would be 1 million, and by the third move, 1 billion.

That kind of "combinatorial explosion" lead to this phenomenal analysis: that the number of possible unique chess games equals 10¹²⁰.

Must we mention this is an embarrassingly big number? Let's write it out:

1,000,000,000,000,000,000,000,000,000,000,000,000,000,
000,000,000,000,000,000,000,000,000,000,000,000,000,
000,000,000,000,000,000,000,000,000,000,000,000,000,000.

Really?
Numbers like this give any computer the willies -- parallel processor, gallium arsenide, we don't care. Working at 200 million positions per second, Deep Blue would still require a while (an eternity?) to evaluate all possible games: 10¹⁰⁰ years. If that doesn't sound longer than the ads in the last quarter of a Super Bowl, remember that the universe is only about 10¹⁰ years old.

Despite these unfathomably large numbers, in 1957, artificial intelligence pioneers Herbert Simon and Allen Newell predicted that a computer would beat a human at chess within 10 years. The prediction, Simon says, was part of an attempt "to give some feeling for what computers would mean" to society. And it was lumped with three others -- making music, proving a math theorem, and expressing a psychological theory in computer code -- that were accomplished within a decade.

Simon admits "I was a little too far-sighted with chess, but there was no way to do it with machines that were as slow as the ones back then." Slow? It's hard to remember how clanky those early punch-card-reading computers were. In 1958, the first computer able to play chess was an IBM 704 with about one-millionth Deep Blue's capacity.

Slowly but surely
Given that limitation, you can understand why human chess players scoffed at the prospect of being edged by silicon. But over the years, computer-chess fanatics came up with steady improvements. In 1967, a program called Mac Hack started competing successfully in human tournaments. In 1983, a program called Belle attained "expert" status from the United States Chess Federation.

In the mid-1980s, scientists at Carnegie Mellon University started the work that coalesced in Deep Blue, with a Sun workstation that could examine 50,000 positions per second. The project moved to IBM -- known as "Big Blue" for its dominance of the computer industry, and its corporate color -- in 1989.

The version of Deep Blue that Kasparov beat in 1996 had a measly 200 processors. Today's version uses 512 PowerPC processors running at 133 megahertz, combined with a raft of specialized chips designed to evaluate positions and coordinate the flood of data rushing through the machine.

Here's a time line on the development of computer chess.

So how does it work?