Meanwhile, inside the BESM, the current that flowed through our pentode is triggering further currents in other pentodes on the same board, so that the binary 1 it represented is passed along through circuits representing logical operations on that digit. Not very complicated operations, not very abstruse logic. A pentode is plugged in line with another pentode, so both must be on for current to flow through from end to end. This is the logical relationship AND. A pentode is plugged in parallel with another pentode, so the current flows if either is on. This is the logical relationship OR. A pentode is plugged together with a signal inverter, so that the current switches off if it was on, and on if it was off. This is NOT. And that’s all it takes. Wired together in the right order, these are the only moves required to mechanise the whole panoply of reasoning; to set the yes–no picture growing towards the complexity of a Rembrandt in the Hermitage. Sixteen of AND, six of OR and three of NOT, arranged in a branching tree, make this board capable of adding. It can add the 1 in our first pentode to a zero in another pentode, and produce (of course) 1; then add that 1 to another 1 carried over from a previous addition, and produce 0, with an extra 1 to be carried over in turn, down a wire to the circuit board next in the stack, where the next addition is about to commence. 1 plus 0 plus 1 equals 0, carry 1. Of course, Sergei Alexeievich, sitting up late in 1943 manipulating 1s and 0s with a pencil, could do this himself, and operations so much more demanding that the comparison is ridiculous. But he couldn’t do it in one ten-thousandth of a second, and do it again ad infinitum every ten-thousandth of a second. Here’s the power of the machine: that having broken arithmetic down into tiny idiot steps, it can then execute those steps at inhuman speed, forever. Or until a vacuum tube blows./p>
And in fact ten thousand operations per second is no longer so very fast, as these things go. The computer Lebedev is planning will, of course, use the new technology of transistors, and replace all of those red-hot filaments with modest semiconducting nuggets. But even with vacuum tubes, he can build a machine that runs far quicker. He has done it already: the BESM-2 is the cut-down civilian version of the M-20, so called because it works at twenty thousand operations a second. The M-20 has scarcely been glimpsed outside the laboratories run by the Ministry of Middle Machine Building, whose middle machines are all the kind that have a blob of plutonium at their heart, and ride on top of missiles. Virtually the whole production run disappeared at once to the ‘mailbox’ towns you cannot find on any printed map. And, more secretly still, an M-40 exists, and an M-50 too. He built these to act as the brain of the USSR’s embryo missile defence project. They sit at present in an air-conditioned bunker in the Kazakh desert, cabled up to six different radar installations as inputs, and to a ground-to-air rocket battery as an output. Silos over in the Ukraine lob dummy nukes eastward towards the test site. In the slim interval between detecting an incoming round and the moment it becomes too late to intercept it, Lebedev’s computers have to calculate a course for the counter-missiles. Out in the desert a barbed-wire fence marks off a target zone exactly the size and shape of the city of Moscow. For two years, ICBMs dropped unmolested onto this imaginary metropolis; but then the machines started to score hits, putting the rocket streaking up from the ochre scrubland close enough, if it were loaded with its own live warhead, to have engulfed the attacker in a ball of nuclear fire. Lebedev isn’t sure how practical he thinks it would be to defend the real Moscow with nuclear airbursts overhead. But some fine, demanding work has gone into making the system fast enough. Though it is only at the proof-of-concept stage, Khrushchev is already hinting and boasting about it at press conferences. ‘We can shoot down a fly in outer space, you know,’ says Mr K.
Tonight the BESM is calculating neither trajectories nor the destructive details of one of the Ministry’s man-made suns. The stack of thirty-nine circuit boards finishes adding the string of binary digits it is working on; then does it again, and again, and again, because this is a multiplication, which the machine can only achieve by adding the amount it is multiplying, over and over, with lightning stupidity. A little over a hundredth of a second passes. The BESM has a result, and sends it out of the arithmetical processor altogether, a string of thirty-nine 1s and 0s to be parked in a row of the magnetised chunks of ferrite which serve as the BESM’s memory. Another line of ferrite cores gives up another string of thirty-nine 1s and 0s, and sends it back to occupy the processor instead. This is not a number: it is the next line of the program the BESM is running. The first six digits are an instruction, telling the BESM to compare the result it just arrived at with a previous result, stored at an address in the memory indicated by the other thirty-three digits. Laboriously, shuffling numbers and pieces of program in and out of the one place where it can pay attention to them, like a person whose table-lamp illuminates only one little circle of a vast cluttered desk, the BESM discovers that the new number is larger than the old number. Now the program gives the BESM a command which is not absolute, but conditional, and we move from pure arithmetic into something different, into the world of supposition, of hypothesis, of what might be. If – says the program – the new number is larger, then twitch the figures slightly, by one pre-set increment, and go back; go back to an earlier step in the program itself, by fishing out of the line of the program 0sund just here, and then proceed again through every step in between. The faithful idiot complies. Round it goes. It is executing a loop. Another hundredth of a second passes. And round it goes again; around and around. The BESM cycles tirelessly through the same instructions, working the same changes on very slightly different numbers. It will go on cycling round the loop till the comparison comes out differently. Ah: now it has. Meaning has been advanced. The picture has been dumbly refined. A picture of what?
Compared to the BESM’s operations, so rapid and so simple, Sergei Alexievich’s thoughts drift as slow as the blue twists curling up from his Kazbek, and expand in as many directions. He’s reflecting, as he sometimes does, on the frustrations of working for the military. It’s not that he has any problem with what his machines are used for, behind the wall of secrecy. He remembers – his generation can’t forget – the locust advance of the Nazis in 1941 and 1942. He doesn’t grudge any use of his time that helps prevent that terrible devouring from ever happening again. The trouble is that the secrecy is slowing down the technology. His own best work is sequestered. The military tuck it away where it cannot influence the state of the art; and since only a handful of finished machines are ever required, there is never the chance to find out what good things yet undreamed-of might come, if there were just the chance to play with the stupid power he creates. He has to admit, on the other hand, that there are compensations. The military are first in the queue for every scarce resource, and when you work for them, you borrow their standing as the country’s most favoured customer. He smiles to himself, remembering the story his rival Izaak Bruk told him: a beautifully blatant demonstration, from ten years ago, of what military support could mean. Bruk had sent a graduate student to the Svetlana Vacuum Tube Factory in Leningrad to pick over the latest batch of pentodes, for with tube-based logic, the factor limiting the operating speed is the quality of the tubes. Plug-in rig to check the pentodes, letter of introduction – and the chief engineer at Svetlana still sent the boy away with a flea in his ear. But the physicists of the Soviet H-bomb project were clamouring for Bruk’s machine at the Power Engineering Institute to be up and running, and he’d been given a phone number to ring in case of trouble, and a codeword flower to mention. The boy dialled the number. He said, ‘I’m having some trouble buying my, er, tulips.’ In an apartment on the Nevsky Prospekt, opposite a knitwear shop, polite people took down the details, and told the student to wait two days and try again. Two days, because ‘We only act at the level of the Regional Party Committee’, and it would take that long for the twisting of arms to work all the way back down to the Svetlana chief engineer. Sure enough, two days later the reception at the plant was all smiles. The
Svetlana people couldn’t do enough to load the student up with the very best they produced. This too, of course, was a loop; a very characteristic human loop in the Soviet economy. If the signal that a job’s important isn’t strong enough, it can be led away around a circuit of important human beings, each having a little word on the phone to the next, each boosting the signal, till it arrives back where it started, able to trigger action.
The BESM. A picture of what? Of potatoes. The electrons flowing through the vacuum tubes represent digits; and tonight the digits the BESM is processing represent potatoes. Not, of course, potatoes as they are in themselves, the actual tubers, so often frost-damaged or green with age or warty with sprouting tubercles – but potatoes abstracted, potatoes considered as information, travelling into Moscow from 348 delivering units to 215 consuming organisations. The BESM is applying Leonid Vitvich’s mathematics to the task of optimising potato delivery for the Moscow Regional Planning Agency. Seventy-five thousand different variables are involved, subject to 563 different constraints: this problem is out of reach of fingers and slide rules. But thanks to computers, thanks to the BESM’s inhuman patience at iterating approximate answers over and over again, it is a problem that can be solved.
The BESM is using Leonid Vitalevich’s shadow prices to do what a market in potatoes would do in a capitalist country – only better. When a market is matching supply with demand, it is the actual movement of the potatoes themselves from place to place, the actual sale of the potatoes at ever-shifting prices, which negotiates a solution, by trial and error. In the computer, the effect of a possible solution can be assessed without the wasteful real-world to-ing and fro-ing; and because the computer works at the speed of flying electrons, rather than the speed of a trundling vegetable truck, it can explore the whole of the mathematical space of possible solutions, and be sure to find the very best solution there is, instead of settling for the good-enough solution that would be all there was time for, in a working day with potatoes to deliver. You don’t, in fact, have to look as far away as the capitalist countries to find a market for purposes of comparison. There is still a market in potatoes, right here in Moscow: the leftover scrap of capitalism represented by the capital’s collective-farm bazaars, where individual kolkhozniks sell the produce from their private plots. Somehow, in the hardest times, there are always piles of green leeks here, and fat geese, and mushrooms smelling damply of the forest, and potatoes dug that morning; all so expensive you’d only shop here if money was no object, to stock up for a birthday or a wedding party. When the trade is briskest, the recording clerks sally out from the Ministry of Trade’s little booths and walk among the stalls, carefully writing down prices. But how slow it is! How slowly things move, as customers jostle in these triangles of waste ground next to the city’s bus stations and train stations, compared to the ten thousand operations per second of the BESM!
The market’s clock speed is laughable. It computes at the rate of a babushka in a headscarf, laboriously breaking a two-rouble note for change and muttering the numbers under her breath. Its stock arrives one sack or basket at a time, clutched on a peasant lap. It calculates its prices on cardboard, with a stub of pencil. No wonder that Oskar Lange over in Warsaw gleefully calls the marketplace a ‘primitive pre-electronic calculator’. In the age of the vacuum tube, it’s an anachronism, good only for adding a small extra source of high-priced supply to the system, for those moments when the modern channels of distribution can’t quite satisfy every consumer need. And now even that function is becoming obsolete. When Leonid Vitalevich’s program reorganises Moscow’s delivery system, the efficiency gains should fill the state shops with enough cheap potatoes for everyone. Now, as the seconds pass, the BESM is steadily shaving away the average potato delivery distance in the capital. At present, it seems, a spud must travel an average of 68.7 kilometres from cold-store to shop: but in the basement of the Institute of Precise Mechanics it is already clear that 61.3 kilometres is possible, 60.08 kilometres, 59.6 kilometres, and still the program is showing that the optimum has not yet been reached. The shorter the distance, the fresher the potato, the smaller the spoilage: this is the best index of success the programmers can come up with, since price as such is not available to them as a quantity to be minimised. The state selling price of potatoes has been fixed for many years. 57.9 km, 56.88 km. This is very nearly a 20% improvement. Soon Moscow’s potato supply will be 20% ber b. 55.9 km, 54.6 km. It’s a new world.
Ah, thinks Lebedev, the high-ups love this stuff. Always have done, since we turned the very first machine on. Nazarenko, from the Ukrainian Central Committee, came to see it, in the tumbledown monastery building outside Kiev where we built it. ‘Sorcery!’ he said, and winked, as if we’d just shown him the cleverest conjuring trick in the world. Which Lebedev supposes they had, in a way; thrown some wires and some vacuum tubes into a hat, and pulled free brainpower out of it. It was the kind of magic a good materialist could enjoy, certifiably the product of science even if it looked like a wonder out of the old tales. Ask the computer and it would obey, as ready as a genie in a bottle – and as intolerant of badly-framed wishes. At first the politicians had only wanted it to work its magic on weapons. Then, with Stalin dead, and a cautious pragmatism in the air, they’d been willing to see what else it might do; and now ‘cybernetics’ is universally caressed and endorsed, very nearly the official solution to every Soviet problem. Rumour has it that its magic will be brandished in the new Draft Programme for the Party, the document that is going to lay out Khrushchev’s plan for reaching paradise. An obliging genie; a timely genie. If it can shoot down a fly in outer space, it can certainly sort out a few vegetables. And the best of the magic is, for the politicians, that computers promise to lend speed and decisiveness to what the Soviet Union does anyway. They’ll make the plans run quicker. They won’t require the digging up of what has already been achieved, or demand a reconfiguration of the world that might disturb things only just beneath the surface of this kinder time. Lebedev isn’t so sure. He sees that new technology does need new forms of human organisation, to do it justice, to let it work the magic of which it is capable. The moment may come when the choice has to be made to accept the new machines’ powers to disturb, or else to forgo what they can do. He hopes to be ready, with arguments and alliances, if that moment comes. Because, of course, his own choice is already long made.
It’s time to call it a night. He gathers up his papers, slides them into the security briefcase he’ll leave at the security post in the lobby. He cannot find his own work magical, not in any sense of the word that implies mystery. He knows its innards too well. And yet there’s something about the way mute matter mounts up in his machines, and up and up, pattern upon pattern, until it manifests the patterns of thought, which still strikes him with reliable wonder. The earliest computer he built used sound waves echoing through mercury for a memory. The mercury is long gone, except in his imagination: where, with the logic of dreams, he knows his calling is to make thinking pools of quicksilver, with the world reflected in them.
In the bright pool of the BESM, images waver and crease of Moscow’s 348 potato deliverers, its 215 potato consumers. The economists recognise the difficulty of getting a computer model to mirror the world truly. They distinguish between working ot zadachi, ‘from the problem’, and ot fotografii, ‘from the photograph’. It would always be better to be able to work from the problem, and to make direct enquiries into how organisations really function, but usually it is only practical to work from the photograph, and to follow the data the organisations give you. This calculation, alas, is from the photograph. It deals with potato delivery as it has been reported to Leonid Vitalevich and his colleagues. There has been no time to visit the cold-stores, interview the managers, ride on the delivery trucks. But the program should still work. Conditionals, again: it will work, if the figures are reliable. It will work, if it is indeed possible to redirect the flowof potatoes at will in the way that the program dec
ides is efficient. It will work, if the loops by which the program optimises are compatible with the loops in Soviet life that get things done.
It’s a point of view.
Notes – II.2 From the Photograph, 1961
1 But the BESM-2 is hard at work; and so is its designer: for the histories of the BESM and of Sergei Alexeevich Lebedev, see Boris Nikolaevich Malinovsky, Pioneers of Soviet Computing, ed. Anne Fitzpatrick, trans. Emmanuel Aronie, pp. 1–22. Available at www.sovietcomputing.com. See also D.A.Pospelov & Ya. Fet, Essays on the History of Computer Science in Russia (Novosibirsk: Scientific Publication Centre of the RAS, 1998), and the chapter about Lebedev and the very first Soviet computer in Mike Hally, Electronic Brains: Stories from the Dawn of the Computer Age (London: Granta, 2005), pp. 137–60.
Red Plenty Page 14