to home
Japan: Taking Over or Just Catching Up?

Subtitled Why Japan Is Betting Its Future on Artificial Intelligence, this book advances two main arguments about the gap between language describing reality and reality itself. Its author first of all argues that western understanding of the goals of Japan's drive for a "Fifth Generation" of computers has been based on our poor knowledge of Japanese. He sees the chief villains here as Edward Feigenbaum and Pamela McCorduck, authors of the 1983 book The Fifth Generation and ardent advocates of artificial intelligence in its most extreme and allegedly error-prone form. Through such efforts as theirs, the author believes, the West has been persuaded that the Japanese seek new and advanced computers in order to surpass western technology once and for all.

Unger's thesis is that the Japanese drive for supercomputers is in fact motivated by very different goals. Their main concern has been in fact to find a way of manipulating what he regards, whether on the computer or in daily life, as an intrinsically unwieldy language. And herein according to Unger lies the second gap between language and reality, in this case the inability of the Japanese to recognize the inefficiency of their own language, a blindness which has prompted a vain (in both senses) attempt to find a way of perpetuating their system of writing. To achieve this, they have allowed themselves to be sold a bill of goods based on unrealizable goals of a utopian pseudo-scientific effort. And it is this blindness which explains the many postponements the Fifth Generation (ICOT) project has suffered since its inception.

This is heady stuff, and it is no dishonour to be partially or even largely mistaken in an area so new and fraught with sophisticated cross-cuttings of disciplines. Indeed, Dr. Unger's book encompasses far more than merely the problems of Japanese typesetting. The author's detailed understanding of these highly specialized concepts and processes is truly impressive, and he presents a wealth of detail about Japanese kana keyboards and character conversion that will surely serve as a resource for others. His reiteration of John Searle's distinction between "weak AI" (that computers can be useful in modeling and testing how the brain works and applying this knowledge in limited expert system instances) and "strong AI" (roughly, that computer programs and the human brain are ultimately analogous and even interchangeable) is a useful one. Most readers of LM would probably feel that he is correct in siding with the former.

But the battle is not yet over. There are still those who argue that computer science will find a way of totally bypassing language. If they are correct, then all aspects of machine translation might suddenly become ridiculously easy, and translators, interpreters, and old-style linguists would go the way of village scribes, water bearers, and footmen. The new gods of language and meaning would be computer scientists. If they are mistaken, one might expect these newcomers to creep off into a corner and commit electronic seppuku. But more probably the battle will simmer down into an uneasy draw, as is already the case, with computers performing massive and tedious tasks but with "human translators" (yes, one does see the phrase) still crucially needed for any sense of finesse or overview. In this climate, the translator/linguist who can also write simply and describe how words actually work could be in demand. Even if computer people suffered what looked like a final loss, they would still retrench and propose new projects. And these would almost surely be funded—the notion that language can be easily handled is an abiding human belief and will not readily vanish.

I do not agree with Prof. Unger that the Japanese might be better off managing their language with a Latinate alphabet. He quotes Kazuhiro Fuchi, Japan's foremost AI expert, as some kind of evidence for his views, though I believe his words might also prove the opposite:

"When you do things like AI or machine translation, it's inevitable that you have to deal with natural language. The origins of ideas differ depending on which language you take as the base, and if we Japanese are going to do these things, freedom of thought will be impossible unless, of course, the base is Japanese.

"The Japanese language is our own, so you could say it has become a unique reflection of our experiences.....Though English commands prestige, I can see its limitations. As long as Japanese think in a kind of transplanted language system, they can't use their imaginations. If we make Japanese the base, however, I think there's a possibility of our producing a stream of new ideas, because Japanese provides us with maximum freedom.....Now I don't necessarily think of these new ideas as Japanese.....I'm a little unhappy when people talk about "Japanese-style AI" in the narrow sense.

"I myself am aiming for a universal base, but I think that the attitude of `well, since it's universal, you might as well do it in English' is extremely weak, at least as a research strategy for Japanese..."

Despite stereotypes of the Japanese, war memories, or related factors which may affect one's reactions to this quote, it is the subjective experience of many western students who have learned to read oriental characters with any degree of fluency that there is a different sensation involved in reading a character-based language than in reading an alphabet-based one. Chinese, Japanese, and Korean represent the one surviving non-alphabetic means of writing. Even if it could be proved decisively—and I do not believe it can—that alphabets are superior to ideograms in all aspects, it would surely be an enormous mistake if this other writing system were rendered extinct just for the sake of conformity. But these are minor quibbles about an excellent book dealing with complex and timely themes in a thoughtful and responsible manner.

J. Marshall Unger: The Fifth Generation Fallacy: Why Japan is Betting Its Future on Artificial Intelligence, Oxford University Press, New York, Oxford, 1987 (quotation from page 182).