San Francisco — Take one part computer science and mix it liberally with genetic engineering. The result - at least in theory - is the ''biochip,'' a computing device constructed in a way yet to be determined from biological materials such as proteins, enzymes, and DNA.
Today the biochip is just a gleam in the eye of a handful of scientists, but tomorrow it could produce a technology that would make the current ''information revolution'' pale by comparison.
The possibility of constructing computers along biological lines opens up a road that leads potentially (though not inevitably) to the stuff of science fiction - machines that reason and reproduce, computer implants that enable the blind to see or give the human mind a photographic memory or the speed of the computer.
Of course, such devices still belong to the realm of fantasy. No one has yet built even the crudest biological computer. What is practically possible in this field remains unclear. But the explosive growth of both computer science and genetic engineering has created a new potential, and the two fields are beginning to converge.
The time is coming when ''engineers and biologists can begin talking together in meaningful ways,'' comments Rudolfo Llanas, a noted neuroscientist at New York University. ''People are beginning to have very strange, very beautiful ideas.''
The first fruits of this collaboration are ''biosensors,'' devices that use biological materials to monitor the presence of various chemicals and convert this information into electrical signals that can be fed into a computer. These sensors, now in the experimental stage, could be used to provide valuable, on-site information for medicine, efforts to identify environmental contamination, and industrial processes such as fermentation.
While important in its own right, ''biosensor technology could provide a natural basis for biological-style computing devices,'' suggests Joseph L. Higgins of the University of Pennsylvania.
Scientists know it can work, because even the lowliest bacteria process information in extremely sophisticated ways. But despite considerable advances, biologists still don't really know how even the simplest organisms compute. And, though the experts have clear ideas on how to design and build biosensors, they have only the haziest notions of how to arrange the basic building blocks of living systems - proteins, enzymes, and nucleic acids - into information processors.
''Biology is theory-poor,'' laments Eugene Yates, the Lincolnesque director of UCLA's Crump Institute of Medical Engineering. ''As a result, we are reduced to telling Kiplingesque stories, like how the camel got its hump.''
Still, proponents of the biochip can list a number of possible advantages for computers constructed from the carbon compounds of biology rather than etched on silicon. These include:
* Extreme compactness. Despite the dramatic miniaturization of today's electronic circuits, single-celled organisms are much smaller, typically no larger than the space between elements on a computer chip. Silicon technology has run into limitations that make it difficult to shrink circuits much further. But mimicking Mother Nature might result in computers hundreds, even thousands, of times smaller than today's devices.
* Three-dimensional structure. Silicon technology is essentially two-dimensional. Biological systems are intrinsically three-dimensional, and 3-D structures package more power into a given space and make possible more complex architectures.
* Self-assembly and reproduction. Biological molecules assemble themselves in complex structures. ''You take the components for a silicon computer, put them in a box, shake them up, and nothing happens,'' says Raymond Jefferis, an electrical engineer from Widener University in Pennsylvania. ''You take the pieces of a living organism, put them in a box, shake them up, and you have a computer. That certainly would make it easier to build microscopic devices.''
* Implantation. The human body offers a hostile environment for silicon-based products. A processor made of biological materials would presumably be more compatible for medical uses. Also, a biochip might be powered by the same form of chemical energy used by living organisms and so would need no electrical power supply.
* Self-repair. Microscopic biological computers would be vulnerable to damage from radiation and cosmic rays. Microorganisms have developed methods to cope with this problem. Adaptation of their self-repair mechanisms might enable the production of more reliable molecular-size devices than would be possible with conventional techniques.
* Pattern recognition. Living things are very good at recognizing patterns. This skill has proved extremely difficult for computers to duplicate. Despite considerable effort on the part of computer designers, the ability of the machines to interpret visual and speech patterns remains primitive. Possibly a biological computer could perform such tasks more efficiently, especially since the basic elements in biological systems are enzymes, which respond to the pattern of other molecules.
* Adaptation and evolution. Currently computers must be designed, built, and programmed. Biological computers might be trained and selected instead. As conventional computers have become larger and more powerful, the difficulties involved in planning, programming, and debugging them have increased dramatically. If an ability to learn can be incorporated into computer design, there would be tremendous advantages.
Michael Conrad is a slight, intense man with sharp features framed by a close-cropped, raven-black beard. He is a professor at Wayne State University, in Michigan, and he's one of a handful of scientists pondering the conceptual basis of biological computation. For several years he has been using a conventional computer to try to simulate a biological system. He estimates it will be 40 to 50 years before a genuine biological computer will be developed. But development will definitely be worth the trouble, he predicts. And he believes it will be possible to come up with simpler devices, such as a biologically based eye for robots, in the near future.
Conrad envisions a robot eye that would resemble the compound eye of an insect. Each of its cells would consist of three layers. The top layer would contain a light-sensitive chemical, the middle layer an artificial membrane covered with a crystalline network of enzymes that would respond to the photoactive chemical. These enzymes, in turn, would alter the chemical balance in the bottom layer. And this chemical field would be measured by biosensors and converted to electrical impulses. Groups of cells might be tailored so their outputs would guide the movements of various parts of the robot arm.
Still, Mr. Conrad hasn't worked out the details. He doesn't know what chemicals and enzymes he would incorporate in such a device. And he isn't certain that such an eye could be made to recognize visual patterns more efficiently than current systems.
Part of the vagueness that surrounds the entire biochip field arises because efforts to blend computer science and biology are so new. But part is due to a basic difference in the way biological systems and conventional computers process information.
Today's conventional computer is, at heart, a vast network of simple on-off switches, controlled by instructions that manipulate these switches in set patterns. It is a wholly defined system. Although the process can be tedious, computer specialists can determine just how a conventiaonal computer has arrived at a given state. The system can always be tested and verified, at least theoretically.
But biochip proponents, such as Conrad, are talking about replacing these simple switches with far more complex ''intelligent switches'' made of enzymes. As a result, the difficulty of analyzing even a modest network of such computing primitives becomes impossible.
This makes some computer designers anxious. If you don't know how a device gets a given result, how can you rely on it? asks Algirdas Avizienis, chairman of UCLA's Computer Science Department. ''The builder of computer systems needs dependability,'' he adds. ''Testability and verifiability must be inherent. We're already in the strange situation where we study and don't understand the systems we've created ourselves.''
At the opposite end of the spectrum is Richard Feynman, a Nobel Prize-winning physicist from the California Institute of Technology, who isn't fazed by such concern. ''To solve some problems we may be forced to accept answers which we know are only probably correct,'' he says.
This is part and parcel of a more basic question that divides biologists and computer scientists: Do living systems process information in ways that cannot be duplicated by conventional computers? The answer to this question could have a major impact on mankind's self-image, as well as the ultimate potential of the biochip.
There is a basic article of faith among computer scientists that all information processors, from people on down, are theoretically equivalent. But many biologists object that living things are intrinsically different from electronic computing devices.
If the biologists are right, then efforts to create machine intelligence may reach a dead end. If the computer scientists are right, the advantages of creating sophisticated, biologically based computing systems may not be worth the cost of solving the vast technical problems involved.
Despite the tremendous scientific and technical uncertainties, there is a small group of scientist/entrepreneurs who have been lobbying the US Congress for a major national effort to develop the biochip.
One of the most eloquent biochip proponents has been James McAlear, the Harvard-trained president of a small Maryland research firm called EMV Associates. The company has been working on a method for etching electrical circuits on proteins, rather than silicon, a process Dr. McAlear says would produce chips 100,000 times denser than current methods allow. EMV has also been studying a method for biologically attaching electrodes to the brain. In a decade McAlear hopes to have a prototype system that could help the blind to see through use of a miniature TV camera connected to electrodes implanted in the brain.
But a recent conference of independent scientists, convened by the National Science Foundation, concluded that the time was not yet ripe for such a national biochip effort. This meeting was organized late last month by UCLA's Crump Institute. Its director, Dr. Yates, summarized the verdict of the scientists involved: ''There is no clear technological imperative behind biochips. There is some kind of challenge in biology, but we're not yet sure quite what that is.''