Nine months ago I bought a new computer. Looking at today's newspaper, I see that my machine is totally obsolete. For the price that I paid I can now buy something with twice the speed and twice the storage. Not only that, all the newest programs absolutely require that speed and storage if I am to run them.

Frustrating, eh? You buy a state-of-the-art machine, and before you can turn around it's become a dinosaur. We are in a real Red Queen's race, just like in Alice in Wonderland, where you had to run as fast as you could to stay where you were.

You have to keep buying more and more powerful equipment, merely in order to do what you did last year. Meanwhile, the programs become more complex, with bells and whistles and "improved features" that I never knew I needed and am still unpersuaded that I do. I guess that I'm an anachronism, because I'm writing this in WordPerfect 5.1 under DOS, and have no desire for an "improved" word processor. On the other hand, do you like to learn new hardware and software every year or two? Do you think that anybody over the age of 35 welcomes such rapid changes?

Well, perhaps we are close to the end of the process. Surely at some point, computers must reach a limit of speed. Are we nearly there yet?

One answer comes by looking at switching speeds. Today's fastest computer can perform a trillion (a million million) operations a second. Fifty years ago, the best computer was a million times slower than this. If we had seen the same increase in automobile speeds, you could drive from New York to Los Angeles in one-fifth of a second. (Just don't blink on the way.)

A factor of a million in half a century is amazing, but there has to be an ultimate physical limit on the speed of switching, the fundamental element of logical decision-making on which all our computers rely. We can estimate that limit as follows: At a thousand trillion operations a second, the energy needed for switching is enough to tear electrons out of atoms. At that point, the computer chips will disintegrate.

This might make you conclude that we'll never have computers more than a thousand times as fast as they are now. However, there is a loophole. Our speed limit assumes that all operations must be done by the computer serially, one after another. But many of today's fastest machines employ parallel operation, in which many calculations are done at the same time. It's the modern equivalent of the proverb, "Many hands make light work." If a calculation doesn't depend on the answer of another calculation, we can do them simultaneously.

There are machines today that employ this principle to perform more a hundred thousand operations in parallel. This pushes the potential speed of operation of a computer up by the same factor. Putting together higher switching speeds with parallel operations, future computers could be up to a hundred million times as fast.

Are we now at some ultimate limit? Far from it. The human brain does its thinking with "wetware," neurons that operate incredibly slowly by computer standards. Our "switching speed," the rate at which the brain's neurons charge and discharge, is no more than a thousand operations a second. Yet we are able to see and recognize a friend's face in a tenth of a second, less than one hundred of our "wetware cycles." Computers require hundreds of millions of operations to do the same job of recognition, and they do it far less well. Some huge ability to do parallel calculations is present in human (and animal) brains. It shouldn't be long before we understand how this is done, and can mimic the methods in our computers.

Like it or not, we have to conclude that progress in computer speed and capability is not even close to an end. More probably, it has hardly begun. Electronic computers are barely half a century old. The human brain as a thinking device has been in development for at least a million years. That's twenty thousand times as long. It would need a braver man than I am to speculate on the processing power - and yes, I believe it's fair to call it thinking power - that our computers are going to have, a million or even a thousand years from now.

On the other hand, when computers become really sophisticated we won't have to learn how to deal with new software, or ever-increasing demands for faster hardware and more storage. We won't even know about them. My computer will recognize me as the weak-minded and feebly logical creature that I am, and be shrewd enough to present itself to me as no different from the older, smaller, dimmer machine with which I feel most comfortable.

"Borderlands of Science" is syndicated by: