Review of Archaeology of Algorithmic Artefacts, by David Link

In a world of superabundant information and advanced technology it can be jarring to remember how much knowledge has been and is forever being lost. We suppose that the historical record is accumulating around us like Wikipedia or the Web itself, and then we realize its closer resemblance to Snapchat. The history of technology in some sense presents an extreme version of this loss, since added to the ephemerality of its record is the obsolescence of its forms. Drawings, descriptions, photographs, manuals, and memoirs may persist—if we get lucky—but technologies themselves become outdated, inoperable and unintelligible, if they survive at all. In this regard the history of computing technology may be the sorriest field of all. Hardware obsolesces and disappears and storage media become unreadable, of course, while software, if it survives, is iterated forward, erasing its own tracks along the way. Google’s search algorithm from 2010 has been forever lost to history, but so has Google’s search algorithm from just yesterday afternoon, since the company makes improvements many hundreds of times a year. And let’s face it: even could we capture an algorithm and pin it down for analysis, few among us would have the chops to understand its inner workings and implementation details. If obsolescence breeds unintelligibility, sheer technical complexity leads most of us there by another, more direct route.

David Link’s Archaeology of Algorithmic Artefacts inhabits these dilemmas. His inspiration is to head back to the earliest history of computing technology, to Alan Turing and the University of Manchester in the 1940s and 1950s, while also providing a few key episodes toward an extended history of algorithmic thought. At least at the beginning, readers can hope, computing technology was not as complicated as it later became, while the emergence of what we now call “coding” effectively began amid the much storied decoding work done at Bletchley Park during the Second World War. The relevant historical record is limited, yes, and none of the early machines—like wartime Colossus (a special-purpose machine) or the Manchester Mark I computer—have survived. Link’s project must proceed at both the theoretical and practical levels, he explains, since the only way to really understand the earliest computer programs is to try to reconstruct them so they can run—hopefully—on software emulations of the earliest computers. In combining theoretical and practical inquiry, Link’s aim is nothing less than the introduction of a whole new discipline—the eponymous archaeology—and in this he succeeds masterfully, arriving at his own synthesis of software and platform studies that has German-media-studies flair, cryptanalyst rigor, and an evident commitment to archive-based historical inquiry.

Archaeology of Algorithmic Artefacts collects seven previously published chapters, which appeared between 2005 and 2013. Taken together they represent a remarkable body of work, here prefaced by three pages of introduction and rounded out at the end with one additional chapter.¹ The unnumbered chapters appear together in this Univocal edition without a bibliography or (shamefully) an index, and thus each can be taken as standalone publications, yet the accompanying footnotes are all numbered consecutively from 1 to 527 across the whole book, and some effort has obviously been made to shape the collection into a monograph. Perhaps it stands to reason that this “archaeology of algorithmic artefacts” is itself a zany, enigmatic artifact, an elaborate puzzle box filled with compelling pieces to challenge, delight, and frustrate its readers.

The fourth chapter presents a “Prolegomena to an Archaeology of Algorithmic Artefacts,” so it’s a good place to start. Here Link defines what he means by “artefacts” and ponders invention, making the case for archaeology as the appropriate mode of inquiry, at least metaphorically, since the discipline of archaeology allows “the mixing of narration and practical reconstruction, of confirmed fragments and deducto-speculative complements” (110). Link’s theory of technology and technological change is elaborated here partly in reference to a mysterious automaton, Odradek, described in a short story by Franz Kafka (“The House Father’s Concern,” 1919), and partly in reference to the Williams tube, the first reliable random-access memory device, developed at the University of Manchester in the late 1940s and used for a short period before magnetic or “core” memory became the standard. (The Manchester RAM story picks up from elements in the two preceding chapters and returns later in the book.) Kafka describes Odradek as an assemblage of a spool, a crossbar, and a rod, and Link puzzles out some different ideas of what the imaginary Odradek could have looked like. He then freely associates an array of similar-looking components and technologies: a patented winding machine for thread, an early electrical demonstration device (which Kafka might even have seen), and a basket shaped radio antenna (and Kafka’s fiancé worked for a company that eventually sold radios!).

This appeal to “schematic forms” (92)—a repertoire of visible shapes that recur across the history of technology—will remind readers of Siegfried Giedion at least,² although it is less important in the long run than Link’s definition of artefacts as “things that have been skillfully produced to serve a certain purpose.” Unlike the purposeless Odradek, artefacts are objects with functions, and they also possess a “symbolic” layer in the sense that the desired function is sutured to the idea of its own implementation expressed in a specific technical form or arrangement of component elements (82). Technology is a kind of knowledge, the new built upon, and out of, the old. So the light bulb—McLuhan’s medium-as-message par excellence—depended upon glassblowing and the vacuum pump. Once invented, it could be “reinterpreted experimentally,” as a diode then as triode or Audion (amplifier tube), becoming more and more complex, differently functional and opaque, its operation and effects removed from direct observation by the eye (84). The bulb is a good technology to think with here, because in its reinterpreted form as the cathode ray tube (CRT), Frederic Williams and Tom Kilburn adapted it to serve as random access memory. The Ferranti Mark I computers, described elsewhere in the book, had CRT screens that functioned both as memory and display.

Bletchley Park’s Colossus machine was destroyed in order to comply with Britain’s Official Secrets Act, but the early Manchester computers were lost to history literally because of their foundational role. Each new and better machine was built up partly out of the old. The onrush of technological progress reinterprets artifacts in a process akin to poiesis (106), while all that gets left behind are paper records and retired engineers. And these can prove sorry substitutes for the technical knowledge lost. Like an archeologist working with a few scattered potshards, the historian must speculate painstakingly from fragments. It’s easy to stop short or go too far and get things wrong. Link’s prolegomena gives one final example to cinch his case for an archaeology of algorithmic technology: he points to historian David Kahn’s treatment of the “bomba kryptologiczna” machines built by Polish cryptanalysts to crack the Enigma code early in WWII, and he complains that Kahn’s account “strangely teems with errors” (108).³ Though Kahn’s book “remains an excellent historical survey,” he gets the technology wrong, because he hasn’t experimented with the machines himself, “either by rebuilding them or by simulating them in software” (109).

The chapter that follows Link’s prolegomena is in fact his reconstruction of the bomba, simulated on computer and described on paper. It’s an exercise that was originally published in the journal Cryptologia and remains written for specialists. Readers will need a sharp pencil to follow along with Link’s thinking here, and—ironically—they will also have to do some background reading in David Kahn or elsewhere to understand the code breaking context, how the Enigma code worked early in the war, the kind of encryption involved, and the steps taken to break Enigma. Link launches his readers straight into “rotor order, ring settings, and Steckers” (113). He is an accomplished cryptanalyst and talented programmer, it seems, not always a good explainer. (Stecker is German for plug; the Enigma machine had a plugboard, which was used to add an extra layer of encryption.)

Elsewhere in this volume Link’s reconstructions are more compelling for a non-specialist. Using the Christopher Strachey papers preserved at the Bodleian Library, Link has reconstructed two of Strachey’s early programming projects. One is a game of checkers played against the computer, and the other is a text generator that produces love letters. Dating from 1951 and 1952 respectively, these are among the earliest software applications known, written to run on the first actual (i.e., “universal,” not special-purpose) computers ever built. Link is very invested in “firsts” and calls Strachey’s checkers game “The first usage of a graphic display in a computer programme” (176). He also credits Strachey with having invented/coded the ENTER command, which takes Link’s reconstruction to the crucial junction of software and hardware. A dedicated ENTER key on a computer keyboard was many years in the future, yet it would eventually emerge as a surviving descendant of the teleprinter interface that was so important for Alan Turing’s famous work on computability. (Turing imagines an all but infinite strip of teleprinter paper operated on for an all but infinite amount of time.) “The strength of Turing’s concept [in “On Computable Numbers”] was that his universal machine was independent of any concrete implementation,” writes Link, but implementation is ultimately the point: Even when software operates, “hardware stays there as its Unconscious” (186). This isn’t quite Friedrich Kittler’s “There is no software,” but Link comes to appreciate the always-ness of hardware.⁴

Strachey’s love-letter program could generate billions of different letters using two basic syntactical patterns and its store of “word data” sorted into parts of speech. In “There Must Be an Angel: On the Beginnings of the Arithmetics of Rays,” Link reports that he reconstructed Strachey’s program and ran it on an emulation of the Ferranti Mark I that he created, drawing on material at the Bodleian and other archival materials that are now digitized. The chapter forks from love “regarded as a recombinatory procedure” to offer a theory and history of computer memory. As we know, computing requires the rule-based reduction of mathematical operations to basic symbols, and also—hardware—a memory capable of storage and “forgetting” (64). We can thank cryptography and higher mathematics (including Kurt Gödel and Turing) for the former, according to Link’s account, and he credits Freud’s mystic writing pad with the latter.⁵ The Standard Edition of the Complete Psychological Works of Sigmund Freud, trans. and ed. James Strachey with Anna Freud, Vol. 6 (London: The Hogarth Press, 1961). Link notes that James Strachey was the uncle of the programmer Christopher Strachey.] (Like Turing, Freud imagined an infinite piece of paper before dwelling on the mystic writing pad as a model for the mind.) Freud’s mystic writing pad is just a warm up exercise though, a metaphor, leading Link to “one of the first inventions for the volatile storage of data,” the delay line. He traces the first use of delay lines back to the British television industry before WWII, and then follows the story of mechanical memory into wartime radar technology used to detect enemy aircraft.⁶ Radar systems had to store incoming patterns from moment to moment so that moving objects could become noticeable. Radar doesn’t so much represent reality as it does calculate it (74).

After the war Williams developed the technology further, “and in 1948 was able to represent up to 2048 ‘digits’ on [a CRT ] screen. Hallucinatory signs, which only indicated angels, because there were no more enemies in the skies, thus changed into symbols of nothing; pure signs that could take on any arbitrary meaning” (75-6). What followed was the construction of the first computer, the Manchester “Baby Machine” and then the Manchester Mark I, replaced in 1951 by the Ferranti Mark I. There is a Kittlerian geist here, the tale of electricity developed (via war) as a computational medium, the story of arbitrary dots on a screen. They are universal, in the sense of Turing’s universal machine, so their meaning, like love itself—Link muses via Freud and Goethe—is always a projection (77-8).

It will be clear that Link’s archaeology gains from being both practical and theoretical while focusing on the work of some great mathematicians, engineers, philosophers, scientists, and programmers. It’s an intellectual history of technology, not a sociocultural one by any means, and it delivers its punches by charting unknown and lesser-known waters as well as by drawing dazzling, often unexpected connections. Readers will be put in mind of Kittler and possibly Giedeon, but Siegfried Zielinski certainly. Zielinski’s Deep Time of the Media: Toward an Archeology of Hearing and Seeing by Technical Means (German original 2002; English translation 2006) articulates an archaeological program with which Link’s is a congruent.⁷. (Four chapters of Link’s book originally appeared in Zielinski’s Variantology series.) Unlike Zielinski, Link codes. Like Zielinski, he is out for intellectual adventure, in the very least by radically extending the arc of media history and undermining some of its pieties.

Exhibit A in this regard is Link’s chapter “Scrambling T-R-U-T-H: Rotating Letters as a Form of Thought,” which turns his cryptanalytic skills toward the deep time of algorithms. Instead of the electromechanical code wheels of the Enigma machine, Link here starts with rotating paper wheels or volvelles found in the work of the Majorcan philosopher, Ramon Llull. It seems that Llull may have based his volvelles on an Islamic divination device called the zā’irja(136). You ask the zā’irja a question, and if used correctly it will give the answer. It was a form of horary divination, meaning that the precise moment the question—any question—was asked was important for generating an answer, so the face of the zā’irja resembles a zodiac; it works as a combinatory apparatus and requires an encryption key in verse.

The Islamic intellectual Ibn Khaldûn described and pictured the zā’irja in his fourteenth-century universal history of the world, The Muqaddimah.⁸ (Princeton: Princeton University Press, 2005).] Mainly using Franz Rosenthal’s 1958 translation of the massive Muqaddimah, Link sets out “to reconstruct the routine employed to generate” an answer using it (146). This then is a history that returns algorithms to the Islamic world—the word algorithm, like algebra, is derived from Arabic—as it plunges into Islamic codicology and Arabic as a code/ing language, attempting the decryption of an unintelligible artifact and what remains recorded about its associated practice. This is also a history of algorithms that points emphatically toward the present moment, as machine-learning algorithms acquire predictive—we might say divinatory—capacities.⁹ More than 600 years ago, the zā’irja “provided a functionality which the constructors of search engines, and many others, are longing for today,” a truthful answer to any question (169). The procedures used seem to have been enormously complicated and are difficult to explain, impossible to recreate completely. Link’s archeology “aims at winning back as much ordered structure as possible” (147). Had he won the jackpot and completed his reconstruction, presumably he (and we) would be able to divine the future.

Elsewhere Link’s deep time, long history of algorithmic artifacts is no less fascinating. In one chapter he reads Turing through Hegel and Hegel through Turing to historicize “the mechanization of thought as a means of understanding” (15). And in another he tackles “Andrei Andreyevich Markov’s Mathematisation of Writing.” The Russian mathematician was “the first to develop a complete theory that takes into account the connections between letters” (32), and so-called Markov chains underpin today’s algorithmic analysis of textual data. Like Hegel and Turing, Markov and Saussure are part of the same intellectual adventure, revealing by turns via contrast and comparison. Markov’s statistical analysis of text meets Saussure’s mechanistic thinking about language, and Saussure’s abiding interest in language-as-speech drives Link’s challenging explanation of Markov. What Markov’s analysis ultimately found hidden within text was something which “writing supersedes, in the two-fold sense of destroying and preserving”: he found the mouth (53). Because Markov’s method aimed at how letters sequence, it “determines the degree to which text represents orality” (56) rather than unpronounceable chimeras.

The virtues of Link’s discipline are many: the long historical arc, a refreshing hands-on ethos, archive-based inquiry. Like any self-respecting historian of science, historian of technology, or historian of anything, the archeologist of algorithmic artifacts must be knowledgeable about the rarefied objects of his study. The archaeologist is more adventuresome than the historian, however, less wed to chronological narration and historiographical citation, more prone to deducto-speculative admixture and creative, illuminating connections. If Link himself is something like a mostly-decoded decoder ring, he is also a little reminiscent of Howard Morland, the nuclear disarmament activist who figured it would be useful if he deduced the way to make a hydrogen bomb and published the recipe.¹⁰ Morland’s readers do not—we hope—make bombs, any more than Link’s readers will recreate his recreations. One great benefit of both projects is the way they point their readers toward known unknowns and—better still—get them wondering about all of the unknown unknowns that still await discovery. Plenty of unknowns lie buried. They are secrets—like Google’s algorithm or like the bomb—or else they are just forgotten, made fragmentary and unintelligible partly by the hand of time and partly in the dizzying and eternal accumulation of so many new things to know.

Notes