shelves. Vessels and jars with reagents sparkle in the cupboards, and colorful piles of wires and soldering irons, their points still red and uncovered with scale, wait for me. Apparatus, neatly wrapped in plastic, sit on the counters — and their pointers aren't bent yet and their scales aren't dusty yet. Dictionaries, textbooks, reference books, and monographs are arranged on the bookshelves. And in the middle of the room, glistening in the January sun, stands the TsVM — 12, the automatic digital printer, with lacy, multicolored wires in the crystal unit. Everything is new, unsullied, unscratched, and everything exudes the wise, rational beauty developed by generations of craftsmen and engineers.
How could I not dream? And what if I succeeded? Actually, for myself, my dreams were much more modest: not of a supercomputer that would be smarter than man (in general, I'm not crazy about that idea, even though lama systems technologist), but of a computer that would understand man, the better to do its work. Then that idea seemed possible to me. Indeed, if a computer can exhibit definite behavior based on everything that I tell and show it, and so on, then the problem is solved. That means that it has begun seeing, hearing, and smelling through its sensors in the purely human sense of these words, without quotation marks or explanations. And then its behavior could be adapted for any work or problem — that's why it's a universal computer.
Yes, then in January, it all seemed possible and simple; the sea was only knee — high. Oh, the inspirational quality of new equipment! The fantastic green loops on the screen, the confident hum of the transformers, the crackling of the relays, the blinking of the lights on the panel, the precise movements of the arrows and pointers…. It feels as though you're going to measure everything, conquer it all, do it all, and even an ordinary microscope inspires the confidence that right now (with a magnification of four hundred and double polarized light) you will see something that no one else has ever seen!
Why even talk about it? What researcher hasn't dreamed at the outset of a project, didn't imagine handling the hardest tasks? What researcher hasn't experienced that overwhelming impatience when you're rushing — hurry! hurry! — to finish the boring preparatory work — hurry! hurry! — plot the course of the experiment, and get on with it?
And then. and then the everyday lab worries, the everyday mistakes, the everyday failures break your dream's spirit. And then you're ready to settle for anything, just so that the whole thing wasn't a waste.
That's what happened to me.
Writing about failure is like reliving it. So I'll be brief. The plan was like this: we would plug the 38,000 — cell crystal unit into the TsVM — 12, and everything else would go into the crystal unit's input: the mikes, the smell, moisture and temperature sensors, the tesometric feelers, the photomatrices with a focusing probe, and Monomakh's Crown, to compute the brain's biowaves. The source of external information was me, that is, something moving, noisy, changing shape and its coordinates in space, having temperature and nervous potential. You could hear me, see me, feel me, take my temperature and blood pressure, analyze my breath, even climb into my soul and thoughts — go right ahead! The signals from the sensors would have to feed the crystal unit, stimulating various cells in it; the crystal unit would form and “pack” the signals into logical combinations for the TsVM — 12; the computer would deal with them as though they were usual problems, and produce something meaningful. In order to make it easier for the computer, I programmed all the number — words from A to Z in the computer translation dictionary into its memory bank.
And. nothing. The selsyn motors, whining gently, moved the feeler and lenses when I moved around the room. The control oscilloscopes showed a daisy chain of impulses, which jumped from the crystal unit to the computer. The current flowed. The lights blinked. But during the first month the digital printer didn't stir once to make a single mark on the punched tape.
I punctured the crystal unit with all the sensors. I read poems. I sang. I gestured. I ran and I jumped in front of the lenses. I stripped and dressed. I let the feelers touch me (brr! those cold feelers!). I put on Monomakh's Crown and — O God! — tried to influence it. I was ready for any magic formula.
But the TsVM — 12 could not put out abracadabra; it wasn't made that way. If the problem has a solution, it solves it; if it doesn't, it stops. Judging by the panel lights, something was going on, but every five or six minutes the “stop” signal went on, and I had to press the reset button. And it would begin all over again.
Finally, I started thinking about it. The computer had to be performing arithmetical and logical operations with the impulses from the crystal unit. Otherwise, what else could it be doing? That meant that even after these operations the information was still so raw and contradictory that the computer could not bring the logical ends together. So it would stop! That meant that one cycle in the computer wasn't enough. That meant — and here, as usual in these cases, I was embarrassed for not having thought of it sooner — that meant that I had to arrange for feedback between the computer (from the units where the impulses still were) and the crystal unit! Then the raw material would be inputted into the clever cube, transformed there one more time, and then fed into the computer, and so on, until perfect clarity reigned.
I perked up. Now we were cooking! I can condense the story about how 150 logic cells and dozens of matrices burned out because the TsVM and crystal unit were out of sync (smoke, acrid smells, transistors flaming like bullets in an oven, and me — instead of cutting off the voltage on the panel, I ran for the fire extinguisher on the wall!), and how I got new cells, soldered the transition circuits, and coordinated the cycles of all the units — just the usual difficulties of technical realization. But the important thing was I got the project off the ground.
On February 151 finally heard the long — awaited clatter: the machine printed out a string of numbers on the punched tape. Before deciphering it, I circled the table on which the piece of tape lay, smoked and smiled vaguely. The computer had begun behaving. There it was, the computer's first sentence: “Memory 107 bits.”
It wasn't what I was expecting. That's why I didn't realize right away that the computer “wanted” (I can't write a word like that without quotes) to increase its memory bank.
Actually, it was all very logical. It was receiving complex information that had to be stored somewhere, but the banks were already filled. Increase the memory banks! A commonplace task in building computers.
If it weren't for Alter Abramovich's respect for me, the computer's request would have gone unheeded. But he gave me three cubes of magnetic memory and two of ferroelectric memory. And everything proceeded smoothly: a few days later the TsVM — 12 repeated its demand, and then again and again…. The computer developed serious demands.
What was I feeling then? Satisfaction. Finally something was happening! I tried the results out on my dissertation — to — be. I was a little put off by the fact that the computer was working only for itself.
Then the computer began building itself! Actually, that was logical too; complex information had to be processed by units more complex than the standard ones of the TsVM — 12.
My work load increased. The printer printed out codes and numbers of logic cells, and announced where and how they should be added. At first the computer was satisfied with standard cells. I mounted them on auxiliary panels.
(I'm only beginning to realize it now, but that was precisely the moment, if you look at it academically, that I made a grave methodological error in my work. I should have stopped and figured out just what circuits and logic my complex was building for itself: the sensors, crystal unit and TsVM — 12 with an increased memory. And then, only when I had it figured out, move on. And when you think about it, a computer building itself without being programmed to do so — what a terrific dissertation topic! If I had done it right, I could have gotten a doctorate right there.
But curiosity took over. The complex was obviously straining to develop. But why? To understand man? It didn't look like it. The computer seemed quite satisfied that I understood it and diligently carried out my commands. People make machines for their own aims. But what kind of aims could a machine have? Or maybe it wasn't an aim, but a kind of innate accumulation instinct, which is found in all systems of a certain complexity, be they earthworms or electrical machines? And what limits would the complex reach?
It was then that I let loose the reins — and I still don't know whether that was good or bad….)
In mid — March the computer, which had evidently learned from Monomakh's Crown about the latest developments in electronics, began asking for cryosars and cryotrons, runnel transistors, film circuits, micromatrices…. I had no time for analysis; I was rushing all over the institute and the whole city, wheeling and dealing, lying and cajoling, trying to get my hands on all this chic stuff.
And it was all for nothing. A month later the computer “got bored” with electronics and “took up” chemistry.
Actually, this shouldn't have been unexpected either: the computer had chosen the best way to build itself. After all, chemistry is nature's way. Nature had neither soldering irons nor cranes, nor welders, nor motors, not
