The 16th Annual North American Computer Chess Championship was barely two hours into its first round when a shock wave passed through the hotel ballroom in Denver: One of the world's most powerful chess programs, running on a multimillion-dollar supercomputer, had been upset by an unheralded little program operating on an 8-bit personal computer, the Apple IIe.
That stunning win set the stage for a fundamental upheaval in the world of computer chess. By the end of the tournament two weeks ago, it was clear that the "little guys" -- the micro- and minicomputers -- could stand up to larger machines, including the world's biggest supercomputers, in the exotic calling of electronic chess.
Since 1948, when American engineer Claude Shannon published "Programming a Computer to Play Chess" -- a seminal paper that is the intellectual basis of every computer game ever devised -- programmers and chess buffs have been working to make machines play this human game well. For most of its history, computer chess has been almost solely the province of big "mainframe" computers. The tournament here suggests that this dominance may be at an end; much smaller machines, including the familiar personal computers, are reaching the top of the computer chess ladder.
A "mainframe" computer is a big (enough to fill a room) and expensive machine; this category takes in everything from the $150,000 computer in a corporate headquarters to the $15 million "supercomputers" from outfits like Cray and Amdahl. A "minicomputer" is smaller and cheaper -- ranging from $10,000 to $150,000. The term "microcomputer" is a synonym for the personal computers we all know -- from the $199 Commodore to a $7,500 IBM PC-AT with all the trimmings.
As computer chess aficionado Philip J. Hilts points out, computers don't really know the first thing about chess. They're just dumb machines. But they can play chess if some smart humans give them enough instructions.
Instructing a computer to play good chess involves translating tens of millions of board positions into mathematical terms and showing the computer how to manipulate the mathematics to find the best move. It would seem logical, then, that the mainframes, with their advantage in memory, speed and number-crunching power, always would prevail.
But the folks who program on micros and minis have some advantages. One is time. Anybody who's programming chess moves on a multimillion-dollar computer is likely to be stuck with computer time at odd hours, when the business users aren't around. The smaller machines are so cheap that it's no problem to devote dozens of hours a week to "frivolities" like chess.
In addition, the personal-computer programmers have the strong impetus of potential wealth if they can design a superior program. There's not much of a market, after all, for chess software for a multimillion-dollar Amdahl; but there are millions of Apple owners who might buy a top-flight program.
These factors all played a part in the North American championships in Denver. The pre-tournament favorite was the defending world champion "Cray Blitz," which runs on the four-processor Cray X-MP 48 supercomputer. There were two programs operating on big Amdahl mainframes, and a couple more that used special networks of big computers connected for parallel processing. One of these, "Ostrich," which runs on a combination of eight Data General computers, was another favorite.
As soon as play began, though, it was clear that the "little guys" would be in the running. The Apple program, written by the British firm Intelligent Software, and an IBM-PC program called "SPOC" (Cypress Software, 1450 Koll Circle, San Jose, Calif. 95112) were holding their own.
And on the last day, when the tournament came to one last game, the mighty "Cray Blitz" found itself up against "HiTech," which runs on a common Sun minicomputer modified with an uncommon circuit devoted strictly to chess.
The game was barely past the opening when David began to whomp Goliath. In the end, the mini program scored a convincing win, becoming the first program for smaller computers to snare the national title. It was vivid proof -- as if any were needed -- that the locus of power in the computer world is moving more and more in the direction of smaller machines.