An Unfair Match Between Brains and Computers

If the match were called "unfair," the computer might want to put in a word first. The contest involved go (or wei-ch'i in Chinese), a game of encirclement played with black and white pieces, called "stones," on a 19- by 19-line board, which Man has been playing for thousands of years. The computer only came into existence 30-some years ago, not to mention its learning to play go.

Fortunately for the computer, mankind is generous. As long as the computer, without being given a handicap first, can defeat a first-level go player (nine is the highest level), before the year 2000, it will win the US$1 million prize.

The instigator of this battle between brains and circuit boards is Ying Ch'ang-ch'i, an ROC businessman, avid go fan, and founder of the Ying Ch'ang-ch'i Go Foundation, an organization which he has well funded to promote the game.

It's not to seek out new opponents besides people but to spur the development of computer intelligence that the foundation, together with Multitech Industrial Corp., cast down the gauntlet and invited computers from around the world to come to Taiwan and match "wits."

Although similar contests have been held overseas, the prize money has never been so high. In response, five computer experts from Britain, the Netherlands, and the United States crossed over lands and seas bringing their programs with them, among them the so-called "pioneer of computer go," a fifth-level player from the U.S., Bruce Wilkes. The foreign programmers, along with eight local Chinese, took part in the 1986 International Computer Wei-Ch'i Congress, held in Taipei on November 11th and 12th.

Because there were so many participants, the computers were required to play off among themselves first--in the two categories of 9 × 9 and 19 × 19 grid boards--to determine which would earn the right to vie with the humans.

In each match, the two programmers would type the moves of the opposing computer into their own, which would then display its response on the screen. With each programmer passing on "enemy intelligence" in this way, the battle would continue back and forth until one side raised the white flag and capitulated.

Computers are methodical players and gentlemanly enough never to try to take a move back. It's the humans, programmers and spectators alike, who can't refrain from ill-mannered kibitzing:

"If you'd just kept on where you were, you'd be fine. Why play way the heck over there?"

"Oh my God! How can you play there? It's suicide!"

"This computer has a different operating system than the one I usually use; it doesn't fit my program. . . . Crashed again! The computer's on strike."

"I changed my program last night, and it seems like it's still hung over today--these moves are madness!"

The winner of the computer-vs.-computer competition in both categories was an entry designed by Tu Kuei-ch'ung, a student at the Taiwan Medical College, while the runner-up belonged to Hsu Shun-ch'in, an assistant professor at National Taiwan University, and Liu Tung-yueh, his student. Bruce Wilkes, who explained that his program was a new one lacking "combat experience," wound up with only a third place showing in the 19 ×19 category.

Representing the side of mankind were two first-level boy players: nine-year- old Hsia Hsien-yu and 13-year-old Ku Ch'ung-ming, who, to give Tu Kuei-ch'ung's computer a better chance, spotted it handicaps of seven and 17 stones, respectively.

Nine-year-old Hsia Hsien-yu handled his event, the 9 × 9 match, with ease, polishing off the computer in less than 20 minutes by a margin of 59 stones.

The 19 ×19 match was more of a contest. After the computer had set out its 17 handicap stones, which gave it a heavy advantage right at the start, Ku Ch'ung-ming's expression appeared serious, and a bit uneasy.

The first hand-to-hand skirmishing, in the lower right-hand corner, wound up a draw. As the struggle progressed up the board's right side, the computer cut off the young Ku's forces, prompting the commentator, a sixth-level go master named Li Ching-hsun, to exclaim, "Good move!"

Unfortunately, as the game grew more complex, the computer lost its fighting mettle, began to slip up, and, much to the crowd's disappointment, dropped the match.

The computer's defeat was not unexpected, of course; otherwise, the humans would not have been so confident as to choose two children as their champions and to have granted a handicap on top of that.

The technology used by a computer to play go is a form of "artificial intelligence," which involves enabling computers to think and make decisions like people. Other examples are using a computer to translate, to show on a screen how a hair style would look on a person before it's cut, and to distinguish fake stamps and signatures from genuine ones.

The computer's not a human brain and can't really think, of course; its "decisions" are all determined by the systematic commands of its program and the information stored its memory. But while it may be a bit stupid, its memory is prodigious and its reactions truly "lightning quick." So in general information processing, the human brain is no rival.

But with "go," it's no go. The main reason is the game's complexity. To program the computer how to respond to each of the possible arrangements of black and white stones on the board's 361 spaces would seem a nearly impossible task. Lacking a grasp of the overall situation, the computer often overlooks easy opportunities or rushes headlong into fatal traps, leaving the programmer to stamp his feet in despair. Even the best go players often can't explain why they've made a certain move beyond "experience" or "intuition," and the scarcity of good computer programmers who are also good go players is one reason for the computer's low level of ability at the game.

Another reason is that the computer must play by Man's rules, which means making all its moves in less than an hour total. The 32K microcomputers currently used to play the game are just too small in memory and too slow, "Computers are now at the same level as a seven- or eight-year-old who's played for half a year," Hsu Shun-ch'in admits.

Children grow up, and computers advance in technology. The latest generation of 64K microcomputers have greater memories and faster speeds than their predecessors, and as the technology advances, so will the computer's go-playing skills.

But the road left is a long one. Computers right now are still at the lowest of 25 preliminary grades below the first nine levels of skill. Before they can compete without a handicap against a human player of the first level, experts estimate they'll need at least 30 more years. That will well exceed Ying Ch'ang-ch'i's 2000 A.D. limit.

Who says people aren't smart?

[Picture Caption]

Where the computer will play next is really hard to predict.

Keeping a record of each game can serve as a reference for revising the program.

Tu Kuei-ch'ung's computer (1.) was no match for 13-year-old Ku Ch'ung-ming (r.).

Hsia Hsien-yu is only nine years old, but quite a good player.

A computer programmer has to be able to bear with solitude and frustration.

Keeping a record of each game can serve as a reference for revising the program.

Tu Kuei-ch'ung's computer （1.） was no match for 13-year-old Ku Ch'ung-ming （r.）.

Hsia Hsien-yu is only nine years old, but quite a good player.

A computer programmer has to be able to bear with solitude and frustration.