'Thinking' computers

From the moment Michael Dertouzos enters the room, you have no doubt he is a man at the technological cutting edge - but taking it all in good spirits. ''You must understand as a premise that what we do here is try to stay ahead - at least 10 to 15 years.'' He breaks into a slightly irreverent chuckle that tells you he's exaggerating but still serious.

''Of course, this is a slightly impossible task, for you must combine some very rational behavior with some very creative insanity. If you're too rational, you're only doing tomorrow's job. If you're too creatively irrational, you may find what'll be good 10 years from now, but lose it to impracticality. We're playing this very tight game between creativity and critical assessment.''

Professor Dertouzos has played the game well, or he wouldn't have risen from boyhood in wartime Greece to direct one of the world's slickest computer think tanks, the Laboratory for Computer Science at the Massachusetts Institute of Technology (MIT).

Dertouzos' imagination projects easily into a world where machines ''see'' and ''hear,'' where people talk to computers and have their wishes obeyed, where businesses and homes are linked in a global information marketplace where people buy and sell data like merchandise.

It is also a world in which those who manage information may wield fabulous power. And this is where the Japanese have brought Dertouzos up short. Back in 1980, he and his 300 researchers had been cruising ahead, basking in their privilege of setting the computer agenda. Then, without warning, the Japanese announced their Fifth Generation Computer Project. Shock waves rippled through MIT's Technology Square.

It wasn't so much that the Japanese wanted to coordinate thousands of computers to do heretofore impossible tasks. It wasn't their desire to build computers out of circuitry with a mind-boggling density of 1 million transistors on a silicon chip the size of a pinhead. It wasn't even their plan to make computers that converse with human beings, translate back and forth between Japanese, English, and German, and dish out expert services in law, medicine, and finance.

What really stunned Dertouzos was the fact that Japan's research goals replicated those of his own laboratory almost to a T. On top of that, the Japanese were declaring that they intended to reach those goals first - even by 1990.

Other scientists and industrialists have not been so impressed. But when Professor Dertouzos learned that the Japanese government wanted to use its ''fifth generation'' research to boost Japan's industrial competitiveness worldwide, save energy, and address social problems, he put two and two together: This was nothing short of a quest for global supremacy in the information field. And he and his colleagues are wasting no time trying to rouse Americans to meet the Japanese challenge.

Pausing from his work one day recently, Professor Dertouzos answered questions on the global stakes involved.

How does it feel when a distant island scoops up your own plans?

All of a sudden we were confronted with an effort to leapfrog Western technology, using our best Western ideas to do it - that was a very big shock indeed. Part of my indignation, I admit, hailed from my European sense of fair play: ''Please, gentlemen, don't copy other people's work!'' But my concern ran deeper. I wrote to presidents of the American computer companies. In the two years since, their level of panic has been rising. Now they realize that the Fifth Generation Project is not some casual copy. It aims at worldwide supremacy in the information field.

How can you be so sure?

MITI, the Japanese Ministry of International Trade and Industry which launched the project, has been saying as much. Their 1980 report announcing the Fifth Generation Project said so. We had discussions with - among others - the project's leader, Tokyo University's Tohru Moto-oka here over lunch. Asked if computer supremacy was the goal, he hemmed and hawed, and finally said, ''Ah, yes.'' Two days later a colleague asked him again before a large audience. He hemmed and hawed, and finally said, ''Yes.''

What exactly is the edge the Japanese seek?

In a nutshell: the ability for humans to extend their mental powers by being able to interact more easily with computers. That means making more intelligent computers that interact more easily with people.

Computers already store vast amounts of data. How would the fifth-generation computers go beyond that?

They would still store vast data bases. But storage is not enough. What if someone sent you to the Library of Congress and said, ''Find all prime ministers who faced at least one war in their lifetime''? Think how long that would take you! So we need not only data, but to have it tailored to what humans want to do. This requires what I call progressively more intelligent computer systems, data structured in knowledgeable relationships. The jargon is artificial intelligence, expert systems, knowledge-based systems.

What the Japanese label fifth-generation systems?

Precisely. The first generation used vacuum tubes; the second, individual transistors; third, their densely packed or medium-scale integration; the fourth , very large-scale integration - which means putting a whole processor on a chip. In the fifth generation, expert systems play the key role - computers that talk, translate, deduce, infer, learn, answer queries. If we can get computers to hear and see, the possibilities would be fantastic.

For instance?

Imagine having an automated stenographer where you dictate and have your words instantly transformed into print. In time, you could ask your computer to go further: ''Listen to what this guy says and write me a story about it.'' With a visual link, you could pick up a piece of printed paper, show it to the computer, and say, ''What do you think?'' You could even keep a personal ''bug'' that has both ear and eye, and keeps a record of everything that passes through your hands. Then anytime you could just ask your bug, ''Who was that Monitor reporter I was speaking to three months ago?''

The implications of this sensory computing for service and manufacturing, of course, stagger the imagination. A manager could say to his computer, ''Increase production by 10 percent,'' or, ''Send back batch 32. It's got defects,'' and have it done! The Japanese are even trying to dramatically improve the design of computers by computers. That could put me out of part of my work.

What's the advantage of boosts in computer-aided design?

Imagine yourself the designer of computer chips - the tiny slivers of silicon that enable whole computers to operate. At present you can get 100,000 transistors on a single chip - the equivalent of 5,000 radios. Not bad. But it takes you three years, and you do much of the work by hand. However, with an intelligent design system - ours or theirs will do - you suddenly have a gigantic aid at your side. You can work with the computer as an intelligent helper to manage far more complex designs, designs with 10 or even 100 times more transistors on each chip. And you can do it all much faster. The Japanese target is to be able to design chips with a staggering 1 to 10 million transistors in a single month.

We're talking about the emergence of a whole new world, new modes of production, a new way of life for all of us - changes more sweeping than the Industrial Revolution. Those with intelligent computers could wield awesome leverage on the world scene.

Would it translate into geopolitical supremacy?

It's not the only factor in geopolitical power. There's food. We control that. There are weapons. But information is an instrument like money: It's not the desired tangible good, but it gets you there. Whoever has a grip on information management will direct a lot of where we go and the bridges that get us there.

How did Japan generate its fifth-generation momentum?

Their government, for one thing, did a lot to get them started. Private companies are usually uncomfortable investing in such long-term research. But MITI is making sure that Japan takes the fifth-generation stride by giving seed money. Its initial $500 million will be matched by $500 million from Japanese computer companies to produce a truly cooperative national research effort in a decade.

In layman's terms, what's the technological leap we need to achieve expert computers?

Let's take an expert system that would have vision. Right now, if I pick up this glass vase and show it to you, you instantly recognize it. Getting a computer to do that is another matter. I could attach a camera, have it measure light reflected from the vase, and get a million light-intensity numbers - that's easy. But now put yourself in that computer's mind: How is it going to figure what it's looking at from a list of a million numbers? Think of all it needs to know - how to interpret, for instance, the reflections coming through the glass, some of which are deceptive. Your knowledge automatically discounts deceptive aspects. The computer would also have to know about mass and gravity. You already understand. You have vast perspective to draw upon. To give a computer even a little of that perspective is monumentally difficult.

But within reach?

We feel we're on the verge of it, if we can perfect the multiprocessor computer. Until now, the basic structure of a computer - the single processor - hasn't changed. It's a single little fellow that sits there clicking away, executing one instruction at a time - you know, ''Add what's in this cell to what's in that cell,'' or ''Put this information onto the screen.'' A home computer operates at a maximum speed of 100,000 commands each second; the world's biggest computers at 100 million per second.

Now 100 million of these little click-clicks per second may sound amazing. But it's not nearly fast enough to make even the most basic identification of a visual scene, or a deeper inference. The only way to reach those speeds is to harness many slaves - that is, many computers working together. We call this a multiprocessor.m It's like a gigantic factory where you bring together thousands of workers who pass along their produce to each other, all coordinated toward a single goal. And to coordinate them is not a simple matter of linking them loosely like you would one home computer to another. It's more like the tightly knit, highly coordinated information exchange that you find in our eye-brain system.

Of course you have a self-interest in getting more US research going. But suppose you were given the green light to tailor an American response to the Japanese challenge. What would it be?

First, stimulate our own long-range research. For this you've got to get government help, since companies are unwilling to invest in research that looks 10 to 15 years ahead. Second, link tax credits to that research. The government would say, ''If you do research that looks 10-15 years ahead, you can get 60 or 80 percent of the cost of this research as a tax credit. Our corporations would respond enthusiastically. To compensate for lost national revenues, bonds could be offered the public, with profits from long-term research going back to them.

Third, get the public to invest in private cooperative research efforts wherever they emerge. One such effort has arisen spontaneously in recent years - the Microelectronics Computer Company consortium. It was spearheaded by my friend Bill Norris, chairman of Control Data in Orlando, Fla. Fourteen computer companies are participating. I'm on the steering board. So far the group has mobilized $40 million for 1983, and hopes to raise $100 million by 1986 with all the participant companies contributing. The consortium will focus on common technologies - like multiprocessor advances - needed by all the companies. Once research reaches a certain level, it would be turned over to companies to let them take it from there. Such cooperation among corporate competitors is a totally new idea.

Are you resigned to simply leave America's competitive, bottom-line thinking intact? Or should we adopt Japan's cooperative model?

I'm a tremendous believer in our free-enterprise system. I've seen other systems, grown up abroad, been through communist repression as a kid, been bombed by the Nazis. I'm convinced that free enterprise is the only system that gives humans full opportunity to expand. Surely we can also make adjustments to keep up with the demands of the time - like, for instance, expanding our long-range vision within a market economy and fostering cooperation through the consortium I mentioned. But I like to think there is room for things to emerge within our free-enterprise system itself.

In some ways, your strategy could become as insular and nationalistic as Japan's. Given today's rampant nationalism, shouldn't we be promoting more global cooperation?

I certainly don't think we should isolate and insulate in response to Japan's project. Let's follow our nation's strengths: Innovate, maintain openness in our universities and in their development of fresh ideas, pursue in a closed way only developments that must remain secret within our companies, but then allow whatever we can to spread through cooperation and openness. These are our fundamental, proven ideals. I think they are contagious strengths. In the spirit of good-natured, free-enterprise competition, we can and should beat the daylights out of our good friends, the Japanese, in the information field and promote the broad welfare at the same time.

What would be the prospect of launching a global consortium, a Sixth Generation Project, to seek planetary computer integration rather than cutthroat , nationalistic competition?

I don't know for sure. But there's no question that information technologies are already having an integrating effect that nobody can stop. Interconnected computers exchange messages that cross national boundaries for business reasons. A worldwide communication network is falling into place at the rate of a million new links a year. Who knows? Maybe we'll see the kinds of consortia you mentioned - Soviets, Chinese, Japanese, Americans harmonizing their individual efforts. For that we'll have to wait and see.