An outfit known as the International Federation for Information Processing held the world chess championships last week, and although there were television lights in the hall, running commentary over a mike and delays in the action, there wasn't a single complaint by the players. As a matter of fact, the players didn't mind at all. They were computers.
The occasion was the second World Computer Chess Championships at Toronto's Hotel Toronto. Sixteen programs from eight countries fought it out while an overflow crowd cheered, International Master David Levy did play-by-play, and the programmers talked and joked even as their computers broke down. Suddenly, chess was in danger of becoming a social event.
But computer chess is more than a sideshow, and when chess masters get over their knee-jerk objections to playing machines, they will discover that computers are the best thing to happen to the game since the Fischer boom. The best programs are no match for masters, but someday they may extend the frontiers of chess knowledge. Freed from stereotyped lines of play, computer programs create bizarre and unexpected variations. They play wild and woolly games, alternately veering into winning and losing positions. And in their present, erratic state, they sometimes make more human errors than humans. A computer once took 86 minutes to ponder a move that was in the book.
The history of chess-playing machines dates back to 1770, when Baron Wolfgang von Kempelen unveiled his automaton at the court of Empress Maria Theresa of Austria. The machine consisted of a desk with a chessboard on top manned by a human-like figure dressed as a Turk. It created an instant sensation. It was not until half a century later, after Napoleon among others had played it, that the Turk was exposed as a fake manipulated by a man inside. The first uninhabited computers that could play an acceptable game of chess came out in the 1950s, but the chess Establishment held them in contempt until 1976, when Northwestern University's CHESS 4.6, the brains of Control Data's beige-colored CYBER 176 computer, beat all-human fields in the Minnesota Open and California B division tournaments.
Now controversy has replaced contempt. Among the spectators in Toronto was Mikhail Botvinnik, 66, the Soviet world champion for 13 years. Botvinnik, who is working on a program of his own, says, "Man is limited, man gets tired, man's program changes very slowly. Computer not tired, has great memory, is very fast." Disagreeing was 92-year-old Edward Lasker, the first master to play against a computer. "Some thoughts you can't program," he said.
Chess is computerized much the same as anything else. All the vital information—material value, board squares, mobility—is given numerical value, with the machine striving for maximum worth. Set to analyze as many as one million variations a move and plan eight moves ahead, the best computers can outplay their programmers, especially in complicated tactical situations. They are weaker at what programmers call "heuristics"—the ability to approximate human thinking—and fall down in such areas as position and spatial relationships.
Just how far computers have come, the overflow crowds in Toronto hoped to discover in the expected showdown between CHESS 4.6, the North American champion, and the Soviets' defending world champion KAISSA, which is named after the goddess of chess. Unfortunately, the match short-circuited. The Soviets took the chance of reprogramming their computer a few weeks before the tournament. They failed to discover a bug, and it cost them an opening-day match with an aggressive Duke University program named, inevitably, DUCHESS. Because the tournament was being played under the Swiss system, with winners playing winners and losers playing losers, the big U.S.-U.S.S.R. match didn't come off. CHESS 4.6 sailed by DUCHESS in the third round and met Ma Bell's own BELLE in the finals.
Perhaps it was just as well. BELLE was every bit as lovable as Star Wars' R2-D2. Asked by its programmer, Ken Thompson, if it would beat 4.6, BELLE spewed out a lengthy answer from the I Ching. An excerpt: "The Prince shoots at a hawk on a high wall. He kills it. Everything serves to further." Alas, 4.6's program was further along. Watching it gobble up BELLE, University of Minnesota computer scientist Warren Stenberg remarked, "The pawn structure looks like a horde of Brazilian army ants." In a special exhibition the following night, CHESS 4.6 beat KAISSA, establishing itself as the world's best, and its young programmers, Larry Atkin (31) and David Slate (32), as the game's new hotshots.
Even announcer Levy was impressed, and he is just a year away from winning a $2,000 bet that he can go a decade without losing to a computer. "I'd need odds the next time," he said. Added Monty Newborn, author of Computer Chess, "Seven years ago the best players came to laugh. This time they came to watch. Seven years from now they'll come to learn."
The United States Chess Federation, which has been slow to acknowledge computer growth, recently voted to allow computers to take part in human tournaments but has not decided whether computers will be allowed to win prize money. If computers ever dominate human play, they may eventually be restricted to their own tournaments. Will they reduce chess to a mathematical formula? No problem yet. There are as many variations to a chess game as there are grains of sand in the universe.
"The first time I saw a computer it was playing in a Massachusetts tournament," says Shelby Lyman, who did television commentary during the Fischer-Spassky series. "I never saw a more depressed person than the guy who was playing it. Finally, the thing got to the end game and lost. I went around saying, 'Boy, is that thing stupid.' That was 1972. Now I'm impressed.
"If computers became better players than people, I would be delighted. I mean, why not? It wouldn't stop humans from playing. Someday computers could help the top masters. Why should we be afraid of such things?"