Progress on the project: an interview with Dr. Kazuhiro Fuchi. David H. Ahl.

In sharp contrast to the ICOT research quarters described by Feigenbaum and McCorduck, 21 months later in April 1984 we found the same area overcrowded and a beehive of activity. The tables were piled high with books, notebooks, and papers. Computer terminals and Prolog workstations were sandwiched between the piles of books and papers. Most of the windows were shaded, probably because the glare interfered with the use of the CRT screens, although peering out at the bleak, gray weather was hardly inspirational. The DEC 20 was on its way out, soon to be replaced by a VAX system and more Prolog workstations.

Although many U.S. researchers in AI question the use of Prolog instead of Lisp, Fuchi is more and more convinced that Prolog was the right choice. He feels it is more suitable for producing the four basic software modules: problem solving and inference, knowledge base management, intelligent programming, and intelligent interface.

I asked how the software was coming along, particulary the modules to perform natural language processing. "So far," Fuchi said, "the focus is mainly on semantics although we are also working on translation between Japanese and English." Success in this area has proved quite elusive to many other researchers over the years. However, the Japanese are concerning themselves only with well-defined technical areas. Hence the size of the dictionary can be limited and there are few ambiguous word meanings.

By April 1984, the 40 researchers were now 42 in number; 3 from MITI, 4 from NTT (the national telephone company), and the remaining 35 from the 8 participating companies. Originally scheduled for three-year periods with the project, more than one-half of the researchers are now planning to stay for a longer period.

As the project nears the end of its third year, support from the government continues to increase. The first year (1981) allocation of *470 million ($2 million) was increased in 1983 to *2.7 billion, and in 1984 to *5.1 billion (over $22 million).

The project has yet to receive any financial support from the participating companies, although they are supporting it with the part-time services of about 100 people.

When the project was first conceived, it did not have the wholehearted support of many companies, even some of those who were furnishing researches. Many saw it as too idealistic. But the past two years have seen a dramatic change in attitude, and today the project has captured the imagination of the computer industry and, indeed, the entire nation. There are still detractors, but that is probably to be expected with any innovative project with a commensurate high level of risk.

His year will see the completion of the initial stage of the project (see chart). Thus far, there have been no spinoff benefits but, said Fuchi, "we don't think of such things. If short-term benefits are what you seek, you don't need a national cooperative project. Individual companies can handle such things."

Many journalists tend to mention the Fifth Generation Project and supercomputers in the same breath. Although the project utilizes some large computers for research, Fuchi feels "The real impact will be on the computers that are readily available to people -- personal computers -- rather than mainframes or supercomputers. The purpose of this project is to develop basic technology. Then, using this technology, you can make big computers and you can make small computers. But," continues Fuchi, "it is more important type."

In his book about the project, Feigenbaum mentions several potential target applications for knowledge based expert systems such as improving the yield of fishing fleets, improving the management of energy, and other similar things. Fuchi, on the other hand, feels that it would be unrealistic to expect such results from the initial ten-year project. He agrees that they expect to penetrate more and more fields, but it will take much longer than ten years.

Twenty-five years ago, the prevailing notion in the field of artificial intelligence was that with more computing speed and more memory, anything could be solved. As this approach appeared less and less promising, researches started to concentrate more on understanding the nature of human thinking and problem-solving processes. Today, researchers have narrowed their goals even more and are focusing on more pragmatic facets of AI, teaching a computer to recognize objects, formulation of theorems in a specific area, and the construction of expert systems. I asked Fuchi about his position in the overall area of AI.

Fuchi confesses to having "a very skeptical opinion of AI for a long time. There were many optimistic people in the AI community. Against that, I was pessimistic and critical. However, I didn't think it was impossible." To realized meaningful goals, however, Fuchi feels it is necessary to utilize many different approaches and methods. For example, says Fuchi, "Is existing computer technology enough? No; it is not that simple." We must also study thought processes and many other things. "If you forget about hardware advancements, you cannot go further. However, to pick only one thing as the important thing is wrong," he says. "Many things are important."

I found it interesting that Fuchi focused on personal computers rather than larger machines for the implementation of the fifth generation technology. In contrast to many other Japanese, Fuchi felt that large companies ought to be changing to more decentralized organizations. "However," he said, "in order to do so, help from many sources is needed." Beyond the desire of managers to decentralize, techonology is also necessary. "Ultimately," he thinks, "this means that the personal computer must become more powerful and widespread."

Likewis, he feels that personal computers will provide increasing support for students at all levels of education. Like some visionary educators in the U.S., Fuchi does not see the benefit of using computers in education to teach students Basic or programming. Rather, he feels a natural language would permit the computer to be used as a far more powerful tool to explore other areas of knowledge.

Throughout our conversation, Fuchi tended to emphasize the long-term point of view in many ways. Indeed, this is typical of the Japanese way of thinking in contrast to the American focus on shorterm goals.

"I understand," he said, "that Americans are studying the ways of Japan. But I think it is not good for them to copy literally." All is not perfect in Japan, he continued, "we have such criticism and problems in our ways."

On the other hand, he emphasized the importance of cooperation in creating the new generation of computers. "Not only Japan," he said, "but America and other countries as well, especially the younger generation, must cooperative in creating the new generation instead of competing for short-term gains." To make a beneficial impact on society and environment, "the entire world must help."