Japan drive to build 'thinking' computer brings out the critics and the competitors
WHEN Kazuhiro Fuchi talks about Japan's program to build computers able to understand human language, he appears assured. The director of the Institute for New Generation Computer Technology (ICOT) does not worry that computers might some day become too independent for human control, as depicted in such science fiction movies as ''2001'' or ''Brainstorm.''Skip to next paragraph
Subscribe Today to the Monitor
''I don't think it will happen in 100 or 200 years. . . . We know such dangers could exist from reading science fiction, so we are trying not to go that way. The more technology advances, the more human beings think. That's why I think it will work out,'' assures Dr. Fuchi.
A more immediate concern for Fuchi is that Japan might lose a global race to design the next - or fifth - generation of computers.
His institute, launched three years ago with about 40 of Japan's top thinkers , was designed to push Japan into the forefront of advanced computer technology. It was also meant to help dispel a long-held image that the Japanese are better at imitating technology than inventing it.
Three years ago, when Fuchi first started, many people were skeptical of the project's viability. No longer. In the United States a group of scientists, such as Stanford professor Edward Feigenbaum, have been pushing ahead with research into ''artificial intelligence'' and the Pentagon has put in $600 million to join the race.
The fifth-generation project has fired up the Europeans as well. Britain has launched a five-year ''Alvey'' program, while another European science group is working on a ''Esprit'' program.
Japan's project is sponsored by the Ministry of International Trade and Industry, academics, the government-run Nippon Telegraph and Telephone Corporation, and Japan's eight leading electronic manufacturers. This $450 million, three-phase plan is expected to stretch over 10 years.
Fuchi refutes persistent rumors that the government is slashing the budget for this project. In the coming year he expects about 5.1 billion yen ($20.7 million), which he considers ''moderate'' given the Japanese government deficit, he says.
So far ICOT has been on target as it nears the end of its initial phase of three years ending March 1985. During this period it has developed a technology base for the future computers that will help it to ''think'' along human lines.
These machines would still process information one step at a time, as in present computers.
So in the intermediate phase the Japanese institute will develop prototype hardware and software for machines that can conduct several processes at the same time, or in ''parallel.''
The final stage will be the development of a prototype machine that will process a complex network of information and which will understand normal language.
Public interest is fixed on the future computer's artificial intelligence - the ability of the machine to make logical inference.
While the present generation of computers can give answers to some questions, it can't tell you how it came up with that answer. But an inference machine will tell you, making such a computer more ''similar'' to the human thought process.
Exactly how clever the the computer will be is difficult to define, but Fuchi likens it broadly to the intelligence level of a primary or even a slower secondary-school child.