Their most secure foothold is in science. They are telling us things about molecular structure and aero-dynamics. In hard-to-quantify sciences like linguistics or psychology they are partly hamstrung; they can't tell the difference between sentences like "Fruit flies like a banana" and "Time flies like an arrow." But they have created their own languages which people must use to communicate with them, and these computer languages throw light on the mysteries of human speech.

THE COMPUTERS that have fore-gone the real world for experimentation in the laboratory have gained exotic accomplishments. A plaque and a trophy sit atop the PDP-6 computer in M.I.T.'s Artificial Intelligence Lab. They were won in chess tournaments. Even psychiatry is tainted. An M.I.T. computer offers comfort to any troubled human. You sit at a typewriter keyboard and inform the machine that your father beats you. "I'm glad you mention your home life," the computer replies politely. "You haven't said much about your mother, for instance."

Computers are finding unlimited opportunity in the arts. Music, for instance. The electric wave which goes from a musical instrument or a recording into the speaker of a sound system can be represented with total accuracy as a sequence of numbers. And since computers can do anything with numbers, they can in principle duplicate not just any sound that the human ear can hear but any sound that can be created. They do it by emitting 20,000 three-digit numbers a second--something no human could ever do--and turning them into an electric wave that can activate a loud-speaker. The computer is a universal instrument limited at present only by humans' knowledge of what numbers will recreate a given sound. As yet the richness of conventional music escapes the computer, but it seems to be only a matter or time before it will be able to sound like the best symphony orchestra in the world or out of it. As for composition, it is theoretically possible that the computer can compose music in a given style, even the style of a particular individual.

Computers have dabbled in choreography, have made sculptures and mobiles. They have written poems, essays and stories (after a fashion) and have generated animations and films. Their graphics are exquisite. A computer's imitation of Mondrian's painting "Composition with Lines" was shown to 100 people in an experiment. Only 28 could correctly identify the computer's picture, and 59 preferred it to Mondrian's.

Computers can do just about everything but leap tall buildings at a single bound, and someday they will be able to do that too. One long-range goal of the technicians in the Artificial Intelligence Lab is to build an "intelligent automaton" that could substitute for men on a Mars expedition. Carrying enough fuel to get to Mars and back seems impossible, so robots will have to go, explore, report back to earth and stay there (safely out of harm's way?). And since there would be a four-minute or worse radio time lag between here and there, communication would be difficult and the robot would have to be able to make his own decisions.

There are basic worries and baroque worries, and a scheme for a robot-astronaut is decidedly baroque. The chief programmer at the Artificial Intelligence Lab, William Henneman, says, "We're still working at things kids have solved by the time they're two years old." What the research on intelligent automata is currently involved in is providing computers with "eyes" and "hands."

The thorniest problem is the eye; for in order to analyze a visual scene a great deal of knowledge about the physical world is needed--knowledge that computers have barely begun to acquire. The method used here is to scan with a camera a scene containing light object on a dark table. The varying light intensity is expressed as logarithms which direct successive scans until a fairly sharp idea of the objects' boundaries are obtained. After many steps an accurate two-dimensional mapping of the scene is completed and translation into three-dimensional models begins. Knowledge from many levels must interact before the computer is ready to put its manipulator into action. The manipulation is also an extremely complicated process which as yet does not yield very dramatic results.

But the results are not important. William Henneman says, "I'm completely unimpressed by facts. I like storing hundreds of thousands of facts in a single sentence." It's the principle of the thing. And the principle behind intelligent automata is no more reassuring than the thought of the ultimate machine. Both suggest power--alien, far-ranging, self-determining, mysterious power. It is a truism that power corrupts men. What will it do to machines?