Computers Cast a Long Shadow on Chessboard

It is a curse upon a man. There is no happiness in chess.

--H.G. Wells

****

Every chess player has a ready, if somewhat defensive, analogy for the day when computer beats man.

A Ferrari can cover a distance faster than Carl Lewis, yet that takes nothing from the runner. A mechanical basketball player with 18-foot arms could beat Michael Jordan, but so what? You don’t invite forklifts to weightlifting competitions, do you?

In New York on Saturday, the world’s best chess player, Garry Kasparov, will begin a six-game rematch with the most powerful chess computer, an IBM machine known as Deep Blue. Last year, their first encounter drew worldwide attention as a computer for the first time won a game against a reigning world champion--even though Kasparov went on to win the match decisively, 4 to 2.

This year, IBM returns with a computer that has double the calculating speed and improved analytical powers. The match is once again likely to become a sort of icon for the epic struggle between man and machine. It will trigger considerable musing about the nature of thought and the dignity of the human species. Most of mankind will root for the man.

For those of us who have devoted embarrassingly large portions of our lives escaping into the absorbing, imaginary world of ideas and combat over the chessboard, the match has a more particular and personal meaning.

There is no reason, even if a computer becomes the best player in the world, that we can’t continue to play on Friday nights at the Pasadena Chess Club or grandmasters can’t gather for annual tournaments in Linares, Spain.

But we sense the shadow cast by chess computers. Most of us who have played over the last few decades already have felt the sting of being beaten by inexpensive computer programs run on average desktop machines. The depth of computing by Deep Blue suggests to some that the game we thought of as involving creativity, strategy, even art, is ultimately as solvable as tick-tack-toe.

Kasparov himself discounts the analogies that chess players find so comforting. “If Deep Blue beats the world champion, it is a different situation to a Ferrari outracing Carl Lewis,” he said recently. “We humans know that there are many animals and machines that are faster, stronger or more agile than we. But none is smarter, more intelligent. In this area, we have enjoyed a monopoly.”

While cars and mechanical basketball players don’t change the way humans run races or shoot 20-footers, computers already have had a huge impact on the way people play chess.

When I was 17 and won the 1972 California Junior Championship, computer chess was in its infancy, and we sneered at the pathetic moves made by the best programs. Today, databases have revolutionized the way players analyze the game and propelled innovations in openings and endings.

The Internet has become a forum for communicating the latest games, as well as a 24-hour chess club for those wishing to hook up with human (or computer) opponents around the world.

‘Deep Blue Is a Cheater’

The contest that begins Saturday will not be a meeting of equals.

Kasparov, a Russian who became world champion at 22 in 1985, is perhaps the greatest chess player ever. He is known for his intimidating ferocity over the board, his will to win, his great tactical and strategic skills. He dizzies human opponents by creating impenetrably complex positions and has a marvelous sense of when to shift from one strategic idea to another. He wants very much to win the $700,000 offered by IBM for beating Deep Blue. (He gets $400,000 if he loses.)

But Kasparov is susceptible to human emotions. He can experience fatigue and make mistakes during the match games, which can last four hours or longer. He has to rely on his excellent, though fallible, memory as he considers possible opening moves. Studies suggest that most players consider two possible moves per second.

In most chess positions, Kasparov will do little calculation, relying primarily on intuition. In rare positions, he will calculate 10 or more moves ahead. He is adaptive and can learn over the course of a match.

Deep Blue is a proprietary program running on a high-end RS/6000 IBM computer weighing 1.4 tons. It has been designed and operated by five top IBM scientists and aided by former U.S. chess champion Joel Benjamin.

The computer, linked by phone to a terminal at the playing site, operates with 32 parallel processors and examines about 200 million positions a second. It feels no pressure and makes no mistakes in any of those calculations, which explains why computers already are supreme in speed chess, when an entire game must be completed in five minutes and tactical errors (by humans) are common.

Deep Blue considers more positions in 10 seconds than Kasparov will think about in his lifetime.

The computer doesn’t learn from mistakes, but the programmers can make adjustments in its approach between games. The computer also has a database with every grandmaster level game for 100 years.

This fact offends many players. “Deep Blue is a cheater,” said Andrew Sacks, a longtime Southern California chess master. “It would be as if I were allowed to play with access to all the latest opening innovations, every line worked out in advance and available to me. All the simple ending positions are programmed in also.”

Sacks, who teaches English at the DeVry Institute in Pomona, said, “Deep Blue doesn’t possess any of the qualities we admire in great chess players: judgment, creativity, intuition, courage.” Even as he insists that the match is meaningless, he admits that the steady conquest of computers bothers him and that he will avidly study the match.

The vast differences between the competitors tell us that Deep Blue isn’t doing anything resembling human thought, a fact that IBM readily concedes.

Mark Bregman, a physicist and general manager of the RS/6000 division, says the company has chosen to invest so much in the project--by some estimates tens of millions of dollars over more than a decade-- because “it is one of the grand challenges that drive your skill. If you’re a mountain climber, you want to be on the tallest mountain. It pushes your limits.”

Asked what the company is learning from Deep Blue, Bregman says the key applications are in “data mining,” looking quickly through huge databases to find patterns. For example, he said, computers have analyzed grocery store sales data and found that “when men buy diapers they frequently buy beer. . . . Maybe that tells you you should locate the diapers near the beer.”

Fortunately for Kasparov, the computer doesn’t understand chess as well it does diapers and beer, although the IBM team has worked this year to improve the program’s heuristics, or rules of thumb. The program will seek to overcome what it lacks in understanding with prodigious data mining.

The problem for Kasparov is that brute-force calculation at some point arrives at moves that look an awful lot like those that result from a deep understanding of the game.

Last year, he played the first game in a macho style as if he wanted to demonstrate that he could even out-calculate a computer. He couldn’t.

“In the first game, Kasparov had no respect for it,” said Jack Peters, chess columnist for The Times. “Then Kasparov was terrified. That won’t happen again.”

How can Kasparov hope to keep up with an opponent that examines so many more possibilities?

The answer lies in the nature of chess, which has long been regarded by computer researchers as an ideal way to test theories of what makes up thought. The rules and object of the game are simple (capture the opponent’s king), and it is played on a finite 64 squares.

In theory, this means chess is solvable. The simpler game of checkers is nearing the point at which every possible move can be calculated by computers.

But, practically, researchers acknowledge that chess remains far out of reach. In most positions there are more than 30 possible moves. For each of those moves there are more than 30 possible responses. The number representing all possible moves is something like 10 followed by 120 zeros, so even at 200 million moves a second, Deep Blue would have to spend eons to work out every possibility.

The vast number of possibilities also suggests diminishing returns from increased computing power. Typically, Deep Blue will be calculating five to six moves ahead. Another doubling of speed probably wouldn’t even add another half move because the number of possible moves increases faster.

In most positions, Deep Blue will waste most of its computing power looking at possibilities that are irrelevant to the game at hand.

The computer thrives in simple endings and in complex tactical positions, both of which lend themselves to solution by brute calculating force. The computer tends to flounder in situations in which the goals are less immediate and subtle judgments are necessary. In the final two games of last year’s match, Kasparov steered toward such positions where strategy was more important and made Deep Blue look foolish--if it’s possible to make a computer look foolish.

If Deep Blue isn’t “thinking” in any reasonable understanding of the idea, and if computers are far from solving the game in a mathematical sense, why should human players be bothered in the least by its possible supremacy?

“I don’t like to talk about it much because it’s so depressing,” said Peters, an international master who lost to Deep Blue’s predecessor, Deep Thought.

Fears About Chess’ Future

A computer victory would devastate chess, though the impact probably would not be felt for a generation, Peters believes.

“I fear that if computers are perceived to play chess better than humans possibly can, then new future Garry Kasparovs won’t study chess,” he said. “Chess requires years of intense study to reach the grandmaster level. It takes a certain personality and a belief that you can do this better than anyone else in the world. If computers are perceived to be best, the people with the biggest egos won’t take it up. We’ll still have chess and still have top players, but not Kasparovs.”

Most top players, including Kasparov, believe that computers eventually will surpass the best human players. Peters isn’t so sure, noting that computer programmers have been making mistaken predictions for decades.

“If Mike Tyson had to box an elephant,” he said, “if the elephant stomped on him once the match would be over. But if Tyson was careful he’d win on points every time.”

So far, computers have not had the impact Peters fears most.

Indeed, databases and the Internet communications have made it possible for young players to get better faster. In late March, Etienne Bacrot of France became the youngest grandmaster ever at the age of 14 years and 2 months. (Bobby Fischer, the longtime record holder, was a grandmaster at 15, but has been superseded by three players in recent years.)

Today, with a database on a Pentium computer a player can examine 50 or a hundred games in the time it once took to play through six games from a book.

For Cyrus Lakdawala, a master ranked among the nation’s top players, technology has been a godsend because there is little international-level chess being played where he lives in San Diego. Through the Internet he downloads the latest games and plays grandmasters around the world.

The access to information has accelerated developments in opening play. In the past, new ideas might take months or years to be disseminated through magazines or books. Now novelties are known everywhere overnight.

“Opening research is getting completely out of control,” said Lakdawala, noting that there are variations of some openings that continue past move 30. (Most games wrap up by move 50.)

He believes this has increased the paranoia among top players, ever more fearful that an opponent will find a trick in an obscure line of their favorite opening. At the top levels, this has forced players to be more unpredictable by venturing a greater variety of openings.

‘No Ego Involved’

The prospect of a computer coming up with stronger moves than any human player may inspire another kind of paranoia. In a high-stakes world championship match, cheating would become possible, in which a player surreptitiously allies himself with a rapidly calculating computer.

Most players don’t bother competing directly with computers. It became a fad in the 1980s to enter new models in tournament competition, but the novelty quickly wore off because humans didn’t like to play against computers.

The only time I played a computer in competition was in 1988 in the San Diego club championship. The newest model of the Fidelity chess computer was entered in the tournament. I avoided complications, won a pawn and eventually the game after 70 moves.

The only pleasure in the victory came from the programmer’s irritation at having to sit through six hours of play.

“There’s no ego involved,” said Lakdawala. “If you win, you’re beating an abstraction. The joy of winning is beating someone who’s alive and suffering.”

He is among the few top players who predicts Deep Blue will beat Kasparov. His reasoning is entirely human, however. “IBM is really cocky this time,” he said.

But Lakdawala also is among those who think such a victory would mean nothing. “You have to know you exist before you count as a chess player,” he said.

As the match nears, the Deep Blue competition has created a frenzy of interest inside the chess world and out. Last year, IBM’s Internet site crashed when more than 5 million people tried to get information about the match. Using a computer not that much different from the Deep Blue hardware to handle the Internet traffic, the company says it won’t let that happen this year. (The Internet site is https://www.chess.ibm.com)

My prediction is that Kasparov will not only win, but win easily. This may not be evident either in the final score or in Kasparov’s comments about how grueling the match was. It is in his interest to bring IBM back for another try next year with another lucrative prize fund.

For a while IBM will keep coming back for the man vs. machine PR. But if the company eventually concludes that Kasparov is pulling an elaborate con, it probably will throw in the towel and focus on selling more RS/6000s to grocery store chains so computers can return to what they are supposed to do: look for connections between diapers and beer.

No other company will want to invest so much money on chess. The current programs will plateau about where they are because of the limitations of computing power in exploring the game.

And human players will continue to dominate.

I hope.

Saylor, an assistant Business editor, is a national master whose team recently won the U.S. amateur championship.