BY SETH PERLOW
One book, written by a computer, could have killed us all.
What do you do when you’re the only country in the world with atomic bombs? You make them much, much bigger. That was the US strategy right after World War II. The Cold War was beginning, and by 1952 the US would have a weapon 690 times as powerful as the one dropped on Hiroshima. To make such a gigantic explosion, the scientists at Los Alamos first needed to create a very strange book, one that proved an important component in the history of computing.
The book is called A Million Random Digits with 100,000 Normal Deviates (1947), and it contains precisely that: a huge table of random digits. It’s about the size of a phonebook. Los Alamos scientists used it to do the calculations necessary for designing thermonuclear weapons. As you might have learned in high school, the motion of subatomic particles is chaotic, so the bomb’s designers had to account for randomness in their calculations. They soon discovered that their random numbers were—well, not random enough. Apparently, picking numbers out of a hat just wasn’t scientific enough. So in 1947 they asked the RAND Corporation, a military think-tank, to produce a very large table of very random digits.
The so-called RAND book illustrates a familiar theme: many of today’s electronics emerged from military research. For example, Norbert Weiner formalized cybernetics based on his efforts to automate antiaircraft guns. Alan Turing developed key principles for computer science, including the Turing test of artificial intelligence, after breaking Nazi codes for the British intelligence service. One scientist involved with A Million Random Digits, John von Neumann, also invented the Von Neumann Architecture, an influential blueprint for computer hardware. The Internet itself began as ARPANET, a project by the Department of Defense group now known as DARPA. (In 1977 the entire Internet looked like this; note the names of universities, corporations, and military bases.) This group also claims to have developed the computer mouse—only partly true—and they’re responsible for those creepy walking robots you might have seen online.
By 1945, academic and corporate research had become integrated with the war effort. The University of California managed Los Alamos, but until the end of the war only one UC official knew its purpose or even which state it was in. The RAND Corporation, short for Research ANd Development, started as a collaboration between the Douglas Aircraft company and the US Air Force; its first publication was the prescient Preliminary Design of an Experimental World-Circling Spaceship (1946). Similar groups included the Institute for Advanced Study, home of Albert Einstein, and AT&T’s Bell Labs, which developed the solid-state transistor during the same year work on the RAND book began. Decades before Google became known for its casual work environments, these groups recognized that a little disorder in the workplace fosters innovation. Researchers were rarely held to specific performance standards and were encouraged to collaborate across disciplinary boundaries. If your company has a foosball table or free beer on Fridays, you might have the military-industrial complex to thank.
A big supply of random digits made it possible to do a new kind of math, called the Monte Carlo method, which simulates the movements of particles in a nuclear reaction. Monte Carlo math uses random sampling to make calculations. Here’s an example to illustrate how it works: if I draw two squiggly shapes on the ground, I can compare their areas by sprinkling grains of rice over them and counting how many grains fall inside each shape. Sprinkling more rice yields more accurate results but requires more tedious calculations, more counting. The scientist Stanislaw Ulam supposedly came up with the Monte Carlo method while sick in bed, playing solitaire. He realized he could figure out the probability of winning a solitaire game by dealing lots of sample games and checking how many were winnable. He purportedly named the method after his uncle, who liked to gamble. Ulam shared his ideas with his Los Alamos colleagues, including John von Neumann and Nicholas Metropolis. Together they formalized the technique. Like von Neumann, Metropolis made other contributions to computer science too, designing and naming the MANIAC computers in the 1950s. The task of computing randomness helped bring them together.
To generate random digits is surprisingly difficult. Computers cannot produce randomness on their own because their design is based on strict logic. Asking a deterministic machine to pick a random number is like asking your microwave to have a favorite color: it just doesn’t compute. Meanwhile any manual process, like flipping a coin or drawing numbers from a hat, would take too long and might not produce truly random results. To solve this problem, scientists designed an electronic “roulette wheel,” which was basically a virtual model of a wheel with a slot for each digit from 0 to 9. They set this wheel to “spin” at 3,000 times per second and then connected it to a random-frequency pulse. With each pulse, the machine would record the position of the wheel at that moment, and they’d have one randomly selected digit.
But where did this random-frequency pulse come from? No one is certain. Given that people at Los Alamos were fiddling with radioactive elements, some have speculated the random pulse came from a Geiger counter pointed at a piece of uranium. Such elements have a steady rate of decay (the half-life), but they emit particles at random intervals, hence the Geiger counter’s weird clicking. It would be quite elegant if a Geiger counter were used for the random pulse. This would mean that the unpredictable subatomic motion the scientists needed random digits to simulate was the very same unpredictable motion scientists used to generate the random digits.
Unfortunately, it probably isn’t true. Because of the bomb, radioactive elements had become precious and would not just be laying around for odd jobs. More likely a kind of vacuum tube provided the random pulses that told the machine when to stop the wheel. The whole apparatus was hooked up to an IBM punch card device and left running. Ironically, the machine had to be reset at least once because it was breaking down. But in this case, “breaking down” means it was becoming too systematic, not random enough. Likewise, the tables were printed directly from the computer printouts because it was feared a human transcriber would introduce errors into this untainted sea of randomness. A newsletter at Los Alamos joked that librarians would shelve the book under “abnormal psychology.” Today the book’s Amazon page offers other hilarious reviews, one of which calls the randomization “sloppy” because “at the lower left and lower right of alternate pages, the number is found to increment directly.”
These days when a computer needs a random number, there are two common possible sources. It can select from a limited table of random digits stored in its memory—a table sometimes copied from A Million Random Digits, which is available gratis online. Or else the computer uses a formula to generate a “pseudorandom” number, one that’s close enough to random for most reasonable purposes but not random enough for advanced applications like designing thermonuclear weapons. New techniques for generating random numbers continue to emerge, some of which look to the natural world for a source of randomness, as the RAND scientists seem to have done. One recent project, called Lavarand, trained digital cameras on a bank of lava lamps and derived random numbers according to the random shapes they make. The tech firm Cloudflare apparently still uses this technique to encrypt a significant portion of internet traffic.
The RAND book represents one big step in a long history of doing math with randomness. The book of digits and the Monte Carlo method have found uses in a range of fields, from thermodynamics and environmental engineering to statistics and finance. A related method, known as the “random walk,” lent its name to a popular book about investing, Burton Malkiel’s 1973 bestseller A Random Walk Down Wall Street. Randomness remains important in a variety of computer applications too. Weather models use randomness to simulate turbulence in the atmosphere. Video games use random numbers to make computerized enemies behave more naturally, less predictably. In fact, that’s how I first heard of A Million Random Digits, in a footnote about randomization in Ian Bogost and Nick Montfort’s excellent Racing the Beam: The Atari Video Computer System.
The RAND book has also interested at least one experimental poet, Jackson Mac Low, who used it to randomize his writing process. Like many other experimental writers, Mac Low employed procedures known as “chance operations,” strategies to make writing a bit more chaotic. Through chance operations, writers hope to minimize the role of personal choice in their work. When the words in a poem appear by chance, not by choice, then perhaps the poem reflects something other than the author’s personal biases. Writers doing chance operations typically use household equipment like dice, a deck of cards, or even words pulled out of a hat. So it’s strange that Mac Low often drew numbers from A Million Random Digits to perform his chance operations. If the point is to disrupt the influence of social and historical contexts, then why choose a piece of equipment with such a grim origin story? Mac Low used the RAND book for a variety of projects during his long career, but it was especially important for his rewritings of texts by the modernist writer Gertrude Stein. As I argue in the second chapter of The Poem Electric, there are surprising resonances between the RAND book and Stein’s work. By using A Million Random Digits to make poetry, however, Mac Low also hoped to redeem the creative energies of the talented scientists who first made this book for such dark purposes.
Seth Perlow is assistant teaching professor of English at Georgetown University. He is author of The Poem Electric: Technology and the American Lyric and edited Gertrude Stein’s Tender Buttons: The Corrected Centennial Edition, which earned a Seal of Approval from the MLA Committee on Scholarly Editions.
“The Poem Electric is a highly original investigation of how ‘electronics enable poets and their readers to animate and rework, rather than reject and surpass, familiar lyric norms.’”
—Marjorie Perloff, author of Radical Artifice and Unoriginal Genius
“Seth Perlow presents a magnificent challenge to the current fashion of ‘big data’ and mathematized literary analysis. The Poem Electric shows how qualitative, lyric intensities embody dispositions that are of indispensable value to us, and which are in productive tension with the world of screens and memes that we inhabit. It represents a wonderful challenge to so many of our assumptions about the value of technology to the humanities and the place of the lyric in our technologized lifeworlds.”
—Joel Nickels, author of World Literature and the Geographies of Resistance