Review of The Fifth Generation Fallacy by Earl Kinmonth

In 1983 The Fifth Generation: Artificial Intelligence and Japan's Computer Challenge to the World by Edward Feigenbaum and Pamela McCorduck (Addison-Wesley) appeared to near hysterical reaction. Several big-ticket US research projects were created to meet the Japanese challenge alleged in this book. Now, moving into its final phase, the Fifth Generation Project (FGP) has produced little of consequence either in terms of hardware or software. Indeed, some are now worried that the gap between expectations (hype) and results may cast a pall over other research projects.

The book under review seems to be an attempt to trade on memories of the stir caused by Feigenbaum and McCorduck while finding a hidden intent behind the FGP, one which makes its failure to achieve a significant fraction of its stated goals irrelevant. Specifically, Unger advances the thesis that "the Japanese committment to strong-AI research is intimately related to the nature of the Japanese writing system." The FGP was really to overcome "the inefficiency of using traditional script in computer environments." (p. 8) By exploring "the borderland where linguistics, Japanese society, and technology meet," (p. 193), Unger purports to find something that even the Japanese backers of the FGP did not realize -- that it was all a device to keep kanji alive in the modern world.

Rather than proving his thesis Unger presents more evi- dence that the Japanese proponents of the FGP took several misunderstood concepts from American research and turned these first into buzz words and then into funding. With little real documentary evidence that kanji processing was a major concern of the FGP, the author resorts to arguing that (a) since kanji are so troublesome, and (b) artificial intelligence, if it could be achieved (Unger rightly thinks it cannot), would solve many of these problems, then (c) the FGP must really have been about the computerization of kanji.

With this approach, much of the book is a diatribe against written Japanese. If only the Japanese would switch to romaji! Although there is much interesting material in the chapters on the mechanics of writing Japanese, the overall tone is in the style of indignant and imperious Westerners who have been advising Japanese on this or that point of their society for more than a century in the let- ters to the editor sections of English language newspapers.

Perhaps kanji do require more time to learn and make literacy more difficult. Nevertheless, some of the cultures with the highest illiteracy rates have exceptionally simple writing systems. Possibly the very difficulty of kanji impart a prestige to literacy that more than offsets the extra time. Perhaps the extra effort required contributes to lower levels of delinquency. In any event, if one is going to explore the linkages of language, society, and techno- logy, these and other questions need to be considered.

Other long sections of the book deal with the mechanics of transcribing Japanese by computer including a dozen pages of keyboard layouts for Japanese word processors, material that would be tedious even in a technical publication. Implicitly assuming the need for a major break through, Unger underestimates the progress made through incremental change. My own Japanese word processor, purchased in 1987 for Y=100,000 offers all of the capabilities that Unger, writing in 1985, could find only in machines costing several million yen. This is less the result of major developments than of fine tuning existing technology. Nevertheless, the cummulative result is impressive.

Overall this book is a multi-dimensional failure. The technical sections are out of date. The linguistic sections are exaggerated and petulant. Exploration of the mechanism by which Japan stumbled into the FGP and certain American academic entrepreneurs set up a Japanese boogeyman to get funding for their own projects never goes beyond the superficial.

☆ See all reviews by Earl Kinmonth.