A JAPANESE `FLOP' THAT BECAME A LAUNCHING PAD
Remember Japan's Fifth Generation Computer Project? In 1981, when the powerful Ministry of International Trade & Industry (MITI) announced this ambitious research effort to put Japan in the forefront of artificial intelligence, shudders ran through boardrooms in the West. Support surged for the EC's ESPRIT electronics research effort. Britain convened the Alvey Committee to map a response. And in the U.S., the MCC consortium of 21 companies sprang up to preserve America's lead in computers. Unless action was taken, warned Michael Dertouzos, director of Massachusetts Institute of Technology's Laboratory for Computer Science, the U.S. computer industry could easily find itself in the same position vis-a-vis Japan as Detroit.
That response now seems overblown. After all, the Fifth Generation hasn't produced computers that "think." Nor has it given Japan the edge in "parallel" systems using hundreds of microprocessors for attacking giant computer problems. Those were top priorities for MITI.
But in failing to achieve those goals, Japan scored an important success: The Fifth Generation forged a new computer-research infrastructure in Japan that brought together scientists in business, government, and academe. And it trained a whole generation in some of the toughest disciplines of artificial intelligence programming (chart).
Now, with the $415 million project winding down, MITI will try to use it as a launching pad for more research. On June 1, the Institute for New Generation Computer Technology (ICOT), the project's coordinating body, will lay the Fifth Generation to rest at its final conference. But MITI will spend $770 million over 10 years on a follow-on effort called the Real-World Computing Project. This time, the focus is on "neural network" systems, patterned on the human brain, that can learn and discern patterns in real-world situations where data is incomplete or where the number of variables is vast--as in macroeconomics or social behavior.
That's where the legacy of the Fifth Generation may help. Before ICOT, Japan had nothing to rival computer laboratories such as those at MIT, Stanford, and Carnegie Mellon, which drew funds from the U.S. Defense Dept.'s Advanced Research Projects Agency (DARPA). "Everything America has in computers came out of a small group of dreamers supported by DARPA," says John Gage, Sun Microsystems Inc.'s director of science. "Building a similar community was vital for Japan."
A RICH LEGACY. Back in 1982, "we were starting from scratch," says Shunichi Uchida, manager of ICOT's Research Dept. Now, says ICOT watcher Ross A. Overbeek, a scientist at Argonne National Laboratory in Illinois, Japan has a competitive intellectual infrastructure. "The big Japanese companies are seeded with people who know the merits and the drawbacks of the technology." One such Fifth-Generation veteran is Akihiko Konagaya, now an NEC Corp. research manager. He worked with ICOT from the start, developing genetic analysis applications for KL1, ICOT's parallel programming language. The machine it runs on will be shown at the conference. "If ICOT hadn't defined the goals and provided funding," says Konagaya, "few companies would have supported this."
Even as they acknowledge the Fifth-Generation legacy, many Japanese are looking at why the project fell short. Most obviously, rapid microprocessor advances in the U.S. left its massively parallel computer design in the dust. By the late 1980s, such ICOT members as NEC, Fujitsu, and Hitachi saw their chips couldn't match those of the Americans. So they stopped pursuing development on Fifth Generation computers.
Now, they are gearing up for the Real-World project. They may fail miserably. But nonetheless, there is again reason for the West to pay close attention. "The people you see in the companies now, the ones working on the advanced stuff, they are the Fifth Generation," says MIT's Dertouzos. And now, they're ready for the Real World.Neil Gross in Tokyo