Harnessing The Speed of Light

By THOMAS McCARROLL

When Alan Huang revealed his plans to build an optical computer, most of his * fellow scientists dismissed the idea as hopelessly quixotic. It was impractical, if not impossible, they said, to create a general-purpose computer that could use pulses of light rather than electrical signals to process data. During one of Huang's lectures on the subject, a third of the audience walked out. At another talk, some of the scientists in attendance laughed and heckled the researcher, calling him a quack and a dreamer. Recalls the 41-year-old engineer at AT&T Bell Laboratories: "I began to have computer nightmares, but I never doubted that it could be done. I wanted the last laugh."

That was several years ago. Few of the doubters were smirking last week when Huang and AT&T unveiled an experimental computing machine based on optics rather than electrons, the first of its kind. The device -- a crudely configured collection of lasers, lenses and prisms -- could serve as the basis for future optical computers 100 to 1,000 times as powerful as today's most potent supercomputers. The potential applications are stunning: robots that can see; computers that can design aircraft from scratch; processors that can swiftly convert spoken words into written text and vice versa. Such practical optical computers are still years -- some would say light-years -- away. Yet many scientists are already predicting that the device will have an impact similar to that of the integrated circuit, which made small personal computers possible. David Casasent, director of Carnegie Mellon University's Center for Optical Computing, calls Huang's work "an important first step" that has "advanced the clock" of the new technology.

Photons, the basic unit of light beams, can in theory be much better than electrons for moving signals through a computer. For one thing, photons can travel about ten times as fast as electrons. And while electrons react with one another, beams of photons, which have no mass or charge, can cross through one another without interference. Thus while electrons must be confined to guide wires, photons can move in free space. This could open the door to radically new and different computer designs, including so-called parallel processors that could work on more than one problem at a time instead of one after another, as today's serial computers do.

But harnessing the computing power of light has proved to be a daunting challenge. The earliest attempts to build an optical computer date back to the late 1950s, when researchers experimented with mercury-arc lamps and even sunlight. Not much happened until the early 1960s brought the invention of lasers, devices that could concentrate light into powerful, high-precision beams. IBM spent four years and $100 million trying to develop a machine that could use laser beams to operate the multiple "on-off" switches that are the heart of all computers. Unfortunately, the switching operations required too much energy, and the devices often overheated. Eventually the company virtually abandoned the project as unfeasible.

The field of optical computing faded into relative obscurity, but it was revived in 1986 by a breakthrough at AT&T Bell Labs. Research scientist David Miller developed the world's tiniest optical switch, a thin chip that in its latest version measures no more than 10 micrometers (0.00004 in.) on a side. Made of advanced synthetic materials, the device can turn on and off a billion times a second without overheating.

Miller's switches became the building blocks for Huang's optical processor, which took five years to develop. His team finished construction around Christmas but did not get the machine to work until last month. The device is far cruder than even the most basic computers: it has no permanent memory, and the only function it can perform is counting simple numbers. Just a small fraction of the thousands of switches are connected. Nonetheless, Huang insists, the machine proves that his principle works. He thinks computer makers will soon replace wiring inside their machines with optical circuits. By 1995, he contends, some 30% of supercomputers will use optical interconnections.

Huang has not convinced everyone, however. Says one scientist: "Huang is like the boy who cried wolf. He's been promising an optical computer for years, and he's still promising. I'm waiting for him to prove that it's practical rather than it's possible." Others are skeptical that optics can compete with electronic computers. Says Bernard Soffer, senior scientist at Hughes Aircraft Research: "Optical computers would have to be ten to 100 times better than electronic ones to justify retooling." Even enthusiasts are guarded. Says optical-computing pioneer Joseph Goodman, a Stanford electrical- engineerin g professor who was once Huang's teacher: "The first commercial general-purpose optical computer will appear between the year 2000 and infinity, and it may be closer to infinity."

When it finally does appear, it may not be American. A group of 13 Japanese companies, including Mitsubishi and Nippon Electric, has teamed with the government's Ministry of International Trade and Industry to launch a ten-year optical-research program. Given the Japanese record in electronics, their interest in optical computers may be the best evidence that Huang and AT&T are on to something big.