Literary Theory for Robots by Dennis Yi Tenen review – the deep roots of AI

<span>Author Wycliffe Hill and his ‘plot genie’ device for coming up with storylines.</span><span>Photograph: Bettmann/Bettmann Archive</span>
Author Wycliffe Hill and his ‘plot genie’ device for coming up with storylines.Photograph: Bettmann/Bettmann Archive

Hark. The end is nigh. “In the industrial age, automation came for the shoemaker and the factory-line worker,” writes Dennis Yi Tenen near the start of Literary Theory for Robots. “Today, it has come for the writer, the professor, the physician, the programmer and the attorney.” Like the end-of-the-planet movies that pelted the multiplexes at the turn of the millennium, newspapers and – increasingly – bookshops are awash with economists, futurologists and social semioticians talking up, down and about artificial intelligence. Even Henry Kissinger, in The Age of AI (2021), spoke of “epoch-making transformations” and an imminent “revolution in human affairs”.

Tenen, a tenured professor of English at New York’s Columbia University, isn’t nearly as apocalyptic as he initially makes out. His is an oddly titled book – do robots need literary theory? Are we the robots? – that has little in common with the techno-theory of writers such as Friedrich Kittler, Donna Haraway and N Katherine Hayles. For the most part, it’s a call for rhetorical de-escalation. Relax, he says, machines and literature go back a long way; his goal is to reconstruct “the modern chatbot from parts found on the workbench of history” using “strings of anecdote and light philosophical commentary”.

This chatbot backstory begins with Arab philosopher Ibn Khaldun’s 1377 Muqaddimah, which includes a description of “zairajah”, a kind of “letter magic” performed via a sort of horoscope, in which a large circle encloses other circles which, in turn, represent various elements and branches of science. Was this an apparatus for analogical reasoning? For astrological projections? Tenen refers to its “computational” cycles and “procedurally generated text”, and likens it to a “14th-century AI performance”. “Think of the electronic databases working in the background of every hospital as giant zairajah circles, spinning a yarn that connects patients, physicians, pharmacies and insurance companies.”

Tenen is fond of ‘lovely weirdos’ such as the 13th-century Mallorcan philosopher Ramon Llull who informed Francis Bacon

Tenen is fond of what he calls “lovely weirdos”: the 13th-century Mallorcan hermit and philosopher Ramon Llull whose rotating paper charts informed both Francis Bacon and Gottfried Leibniz’s forays into binary code and cipher systems; Georges Polti, a Frenchman whose Thirty-Six Dramatic Situations (1895) offered vivid scenarios (“daring enterprise”, “fatal imprudence”, “conflict with a god”) to inspiration-free writers; Andrey Markov, a Russian mathematician whose groundbreaking work developing models of probability – later known as the Markov chain – stemmed in large part from his time studying linguistic patterns in Alexander Pushkin’s verse-novel Eugene Onegin. Their accomplishments are undeniable, but no biographical information is given to justify the label “weirdo”.

Tenen is more at ease discussing technology than individuals. A 1935 writing manual called The Plot Genie Index came with a wheel called the “Plot Robot” to help its users devise stories. The Qwerty keyboard was designed to slow down typing and make keys less likely to jam. As early as 1959, Bell Telephone Labs patented a device for the Automated Reading of Cursive Scripts. Innovations such as the Random English Sentence Generator were funded in large part by the US army and air force. Boeing drew on Vladimir Propp’s 1928 Morphology of the Folktale to enhance its ability to generate useful reports about unusual aircraft events.

kLiterary Theory for Robots is part of a new series from Norton in which academics are asked to condense complex ideas into small volumes for non-specialist audiences. Tricky: professors build their reputations by writing for their peers rather than for the public. Tenen tries to be peppy and populist, but, like a Ted Talk “thought leader”, ends up sounding ingratiating. One chapter begins “Let me let you in on a little secret”; another is entitled “9 Big Ideas for an Effective Conclusion”. Sentences are peppered with exclamation marks and cringey turns of phrase (“Whoa, these things are old!”). Unsubstantiated claims recur. “Modern humans”, we’re told, treat toasters “with disdain”. Apparently, “Not long ago, one way of appearing smart involved memorizing a bunch of obscure facts”. Today, most of us are still Romantics for whom “there’s just something unsavoury about the thought of individual human genius being diminished by mechanical reproduction”.

A former Microsoft engineer, Tenen has sunny instincts. He believes that machine texts have plenty of “potential for creativity and collaboration”. But he is aware of shadows too. Algorithmic biases, super-scale disinformation, the suspicion that AI “thinks like a state [and] understands like a corporation”. Politics is not his strength, though. Acknowledging that coming labour-market disruption may mean fewer doctors, or even software engineers, he adds, “those that remain will also find their work enriched”. Hoo-bloody-ray.

Tenen doesn’t talk about the death of the author, but he does refer to his own sense of a “receding share of authorial agency”. He means that he uses dictionaries, encyclopedias and search engines – and, when doing so, is drawing on “a shadow team of scholars and engineers” who developed those tools. This, to him, is a revelation: “Intellect requires artifice, and therefore labour.” I don’t know anyone – book historian or general reader – who ever thought otherwise.

Literary Theory for Robots: How Computers Learned to Write by Dennis Yi Tenen is published by Norton (16.99). To support the Guardian and Observer order your copy at Delivery charges may apply.