Azerbaijan International

Autumn 1995 (3.3)
Pages 32-33

Computer Challenges
World's Chess Champion

IBM's Deep Blue to Play Garry Kasparov

by Marcus Hopkins


Man vs. Machine. We've heard it so many times before. But this time it's a struggle between intellects. International Business Machines (IBM) has recently announced that their specially-designed chess computer, Deep Blue, will soon be playing World Chess Champion, Garry Kasparov.

The invitation for the match was initiated by the Association for Computing Machinery (ACM) in commemoration of its 50th anniversary. The event will take place February 10-17, 1996, at the Pennsylvania Convention Center in Philadelphia and continue for six, full-length games. Kasparov, who was born and grew up in Baku, Azerbaijan, has held the World Title since 1985 when, at the age of 22, he became the youngest World Chess Champion ever.

Is it Fair?
For the past several months, there's been a rather heated debate going on among members of chess newsgroups on the Internet, <rec.games. chess>, as to whether such a challenge between man and machine is really even fair. Does any human being have the slightest whisper of a chance to beat a machine such as Deep Blue which, by February, will be able to analyze up to one billion chess moves per second? With such brute force and speed, how could such a machine possibly lose?

"We're not trying to embarrass the world's best chess player," insists Gerald Present, a physicist who works in IBM's Public Relations. "Obviously, Garry wouldn't waste his time and risk his reputation if it weren't a challenge for him and if he didn't think he had a good chance of winning."

According to IBM scientists, Deep Blue is simply a project in computer science that enables them to understand more about a particular kind of computing called "parallel processing", a research effort that IBM has been involved with for the past 15 years. Parallel processing is the technique built in to modern computers enabling them to achieve their highest possible speeds.

Today, IBM is exploring chess, which has long been considered one of the most complex challenges because of the sheer number of possible moves involved in the game. Tomorrow, they hope to apply their technology to such problems as weather forecasting, oil-well drilling, and other tasks that require instant analysis of hundreds of thousands of bits of inexact data.

Kasparov has always insisted that no computer will ever beat him-that there is more to playing chess than mathematical calculations. He has always counted on intuition, profound ideas, spontaneous aggression, and what he calls "art" to win at chess. In the past, he has played computers at "speed chess".

There have been a few five minute "blitz" games and 25 minute "speed" games. Sometimes, he has won; sometimes, not. But he has always insisted that at standard length chess, he can beat any machine. Standard chess allows each participant two hours to complete 40 moves. If the game is not won by then, each player is allowed one more hour to complete 20 moves and this pattern continues until the game is won or drawn. When he's competing against a computer, Garry is convinced the extra time will give him a better chance to win despite the machine's ferocious speed.

Human Strategic Advantage
As Kasparov points out, "Millions of games have been played and thousands of books written, and yet there is no chess formula or method which can guarantee victory. There are no mathematically valid criteria for evaluating even a single move, let alone a position. No genius can master all the opening lines and defenses."

Despite how fast Deep Blue can analyze chess positions, there are certain advantages that favor human beings. When a player starts to think about a move, an evaluative process kicks in, allowing the player to quickly eliminate a whole series of worthless alternatives-moves that he doesn't even have to consider because he knows they're irrelevant. The computer, on the other hand, has to process all choices sequentially without eliminating any possibilities in advance. In other words, there are more complicated mental processes going on inside the human brain than inside even the most sophisticated computer system.

Deep Blue is not designed to imitate the human brain at playing chess according to IBM specialists. "This is not a case of Artificial Intelligence (AI)," observes Present. "Despite incredible advances in science, experts don't really know very much about the way the brain works, and there is only so much insight that a human chess player can provide about how he or she goes about determining which chess piece to move." Psychologists have suggested that playing successful chess may be the product of an innate predisposition for spatial forms allied with an exceptional memory. How else can one explain the phenomena that some people can play simultaneous games blindfolded?

"We've assigned Deep Blue a simple problem of searching," says Chung-Jen Tan, Manager of the Computer-Chess team scientists at IBM's Thomas J. Watson Research Center in Yorktown Heights, New York. "The number of moves that exist on the chess board is estimated to be 10120. The computer's task is to examine all the different possible moves that can be made and then come up with some sort of scheme for 'evaluating' which is best. As such, the machine has to have some sort of quantitative way of determining whether a specific position is good or not. And yes, there are times when a computer can make a stupid, naive move. A computer can only play the way it has been programmed. It's only as 'smart' as the design inside its system. Computers don't think on their own," admitted Tan, who considers himself only an amateur player.

Thinking Ahead
Always the secret in chess is the ability to anticipate an opponent's future moves. The computer has to be programmed to look ahead. The best move may be five or ten steps in advance. Currently, Deep Blue can examine all possibilities for each possible move up to 11 steps ahead. By February, it should be able to "see" considerably further.

Tan admits that all the "kinks" are not worked out of Deep Blue, and that it can't play the perfect game of chess yet. But he expects a close match and is extremely proud that IBM's system was chosen for the match.

Garry has the added advantage before the match of getting a chance to study Deep Blue's public performances against some of the International Grand Masters that it has played. Kasparov uses similar analyses when preparing to play human opponents. "When I know my opponent's patterns and tendencies, I can draw him into traps," he admits.

Kasparov, himself, has already beat Deep Thought, the prototype for Deep Blue in 1989. But the computer's computational capability at that time was only two million positions per second.

History of Chess Machines
The concept for Deep Blue originated from computer scientists, Feng-Hsiung Hsu, Murray S. Campbell, and A. Joseph Hoane, Jr., when they were grad students at Carnegie-Mellon University in Pittsburgh. Hsu wrote his Ph.D. in 1989 about a computer system which he called "Deep Thought" after a science fiction movie. Upon graduation, Hsu was invited to IBM. In 1991, when new hardware was incorporated into the system, IBM, who prides itself as being the "Big Blue", took the opportunity to rename the chess program, "Deep Blue"- "deep" still referring to the layers of calculations and algorithms required in the complex analysis.

The idea for a chess machine originated in 1769 when Baron Wolfgang von Kampelon in Hungary created a wood automaton which he crowned with a turban and, subsequently, dubbed, "The Turk". This device was said to have beat players such as Benjamin Franklin and Napoleon. But the "automaton" was later discovered not to be a machine after all, but a human cramped up inside the box. Serious chess programs and chess computers were first invented and designed in 1950s.

Man is Really the Winner
If Kasparov loses, some chess enthusiasts lament that it will be end of chess. Why play chess if a machine can play it better? IBM doesn't agree nor does Garry.

To IBM it's not a question of the superiority of a machine or a human being. "We're trying to solve a research problem," observes Tan. "We may not succeed in scaling Mt. Everest this time, but if we even get close to the peak, we'll consider it an achievement for mankind. Since human beings are the ones who design complex machines, if Deep Blue wins, it's an indication how far human beings have come in being able to program a machine to do a specific but very complex task. No matter who wins, it's really a victory for humanity."

And for Kasparov, who has been quoted as saying, "In chess, when you lose, you die," a defeat to Deep Blue would come as an incredible blow. Still, it's the psychological interaction with other human beings, not machines, that intrigues him most. "Everest is no less beautiful or alluring to man because an airplane can fly higher. The biggest battle for man will always be against himself."
In the meantime, the winner (be it man or machine) "walks" away with $400,000; while the loser "pockets" the remaining $100,000.

From Azerbaijan International (3.3) Autumn 1995.
© Azerbaijan International 1995. All rights reserved.

Back to Index AI 3.3 (Autumn 1995)
AI Home Page
| Magazine Choice | Topics | Store | Contact us