A byte closer to `thinking' computer. Japanese research group chips away at computer that uses logic
Tokyo — If someone invented a computer that could think, he would bring about a revolution in the entire computer industry. That is what Kazuhiro Fuchi and his 80-odd researchers at ICOT -- the Institute for New Generation Computer Technology -- are up to.
``We want to build a computer that's never been built before . . . to start a current of change that will affect the whole world,'' Dr. Fuchi said in a Monitor interview here. It's an ambitious, and in a sense surprising, project for the Japanese to be working on.
Until now, the Japanese have been celebrated for the superb quality of their workmanship and for their ability to build on and to adapt original ideas that might have been developed elsewhere. ICOT, however, starts from scratch, or as close to scratch as a scientific project can be these days. The immediate goal is to build what Fuchi calls a ``parallel inference machine'' -- a computer that works by logic and inference rather than by a series of consecutive calculations -- as an important step in achieving a thinking computer.
But ``it isn't as if a number of us got together and said, `Right, let's pool the results we've achieved so far and see where we go from there.' '' The project has brought together eight companies plus the government's Electrotechnical Laboratory. But collectively the participants are taking a leap in the dark. ``We may fail,'' Fuchi says, smiling.
He is probably being too modest. In addition to being logical, the computer he's working on would use natural language, the language you and I speak. Existing computers are built on a concept pioneered by John von Neumann and are called sequential because, no matter how fast, they can do but one calculation at a time.
Fuchi, who, soon after graduating from Tokyo University in the 1950s, spent a year at the University of Illinois helping design an early von Neumann-type computer, is now grappling with a totally different concept.
``If A is true, and B is true, then C must also be true.'' This is how inference works, Fuchi explains, and the ICOT team is busy designing a computer that infers, leaving von Neumann far behind.
ICOT was founded in 1982 under the auspices of MITI, Japan's Ministry of International Trade and Industry. MITI has given Fuchi 10 years to perfect his inference machine, with a budget of $400 million. He has a private office for receiving visitors, but prefers to sit in one corner of a large, airy room where his researchers work at desks, beside banks of computers. The atmosphere is relaxed and informal; most of his staff are in their 20s and 30s -- unusual for a Japanese institute.
So far, ICOT's researchers have built a sequential inference machine that is still fairly close to existing computers. The next step, expected in the spring of 1988, is to build the first of what is hoped will be a steadily improving series of parallel inference machines. By 1992, 10 years after beginning the project, a working parallel inference computer should have been perfected.
The United States is also working on a new generation of computers, under the aegis of the Pentagon's DARPA (Defense Advanced Research Projects Agency). DARPA's scientists have the Defense Department as their exclusive client, so that their research focuses on the military aspects of thinking computers -- for instance, how to build a pilotless vehicle that might lie in wait for an enemy tank and destroy it. ICOT's research, on the other hand, focuses on such civilian targets as devising a voice-activated translation machine.
Fuchi wants his institute to be open to the fresh breezes wafting across the world. Under his open-door policy, ICOT welcomes exchanges with counterparts in Europe and North America. ``The British are doing good work in this field -- so are the Swedes. . .,'' Fuchi said.
Is the kind of rote learning that characterizes much of Japanese education enough to produce the original, creative geniuses that Japan needs in increasing number?
``The Japanese system is good at giving students a knowledge of the fundamentals -- say through high school,'' Fuchi said. ``Beyond that, especially in science, there has to be adaptability. You have to be ready to react to all sorts of circumstances . . . Perhaps there's too much uniformity in university education here.''
How about the alarm voiced some times in the US and Europe, that Japan may be about to overtake them in the race to develop the next generation of computers?
``From the very beginning,'' Fuchi said, ``our object has been to help build a new society. From that viewpoint, we want to share the results of this research with everyone. . . . Some Americans seem to be alarmed by what we are doing, but at the level of fellow scientists, I think we are understood. It's only by pooling the contribution of scientists from all over the world that we are going to arrive at the new age. So there's not much sense in being nervous about who is going to win. It makes much more sense for each country to make the contribution it can.''