I first encountered a poem written by a machine eight years ago. I was in a bookshop in Mumbai, participating in a panel discussion called “Imagination Is an Algorithm.” One of my fellow panelists was the Israeli poet Eran Hadas, who was developing software for generating artificial poetry. As Hadas read a selection of machine-written poems, I realized why I had been invited to join the panel: I was an editor of the Penguin Classics series at the time, and the organizers clearly hoped I would come to the defense of old-fashioned analog poetry. As an advocate for Rimbaud, Keats, Shakespeare, and all the flesh-and-blood poets who ever scratched away in a garret, I was expected to protect the honor of the muse Calliope and denounce Hadas’ algorithmic word salads as impostors.
The problem was they weren’t bad. They were minimalist, certainly, and barely punctuated, but they contained evocative images, arresting vocabulary, and fruitful juxtapositions. They sounded like poems. So, why were they also uncanny?
“They’re beautiful, in the way that a waterfall or a flower is beautiful,” I said. What I have since articulated more clearly to myself is that a poem is a thing of beauty, yes, but it is also a form of communication. A poet crystallizes an insight; their readers revive and reimagine it. An understanding is shared between two people across time and space. A poem written by a machine is different. Does it make sense to seek meaning in a poem generated by an algorithm, or is it as one-sided as spotting shapes in the clouds?
Today, nearly four years after the launch of ChatGPT, the situation is more complex because generative AI is becoming so impressively proficient. For example, I just asked ChatGPT to “write a poem in the style of William Carlos Williams.” It gave me this:
“Rain at the Bus Stop”
Rain
holds
for a moment
on the sleeve
of a blue coat —
then
without sound,
joins
the dark pavement
already
full
of sky.
Not bad. It’s increasingly difficult to spot what has been written by a machine, to the extent that the novelist Sarah Hall recently included a “Human Written” badge on the front of her novel Helm (2025), a fair-trade-style assurance of authenticity and provenance. As human creativity faces an existential crisis, what is the future of literature in a post-ChatGPT world?
Literature itself may hold the answer, because our present anxieties have been anticipated and explored in stories for at least 200 years.
In E.T.A. Hoffmann’s 1816 short story “The Sandman,” for example, the student Nathanael falls in love with the beautiful but curiously cold Olimpia, only to discover that she is a clockwork automaton doll: He has misread her fixed gaze as adoration and her lack of speech as modesty. The story is an unsettling portrait of Nathanael’s mental breakdown and a satirical comment on the male gaze, but it also brilliantly captures the uncanny experience of interacting with artificial intelligence. Sigmund Freud used “The Sandman” as a central example in his seminal essay “The Uncanny” (1919), and in 1970, the roboticist Masahiro Mori coined the phrase “uncanny valley” to describe the eerie revulsion we feel toward machines that seem almost human but not quite. Nathanael feels “numb with horror” when he sees Olimpia’s face with “pits of blackness” where the eyes should be; I find myself sliding into the uncanny valley when I read the poem “Rain at the Bus Stop.”
The ethical implications of creating artificial intelligence are perhaps most compellingly explored in Frankenstein (1818) by Mary Shelley. When Victor Frankenstein succeeds in bringing his creature to life, he finds himself in the uncanny valley. The beautiful body parts he has assembled suddenly appear monstrous: “the dream vanished, and breathless horror and disgust filled [his] heart.” Frankenstein’s volte-face anticipates Google vice president Geoffrey Hinton’s 2023 resignation. Hinton’s work in developing artificial neural networks would win him a Nobel Prize in 2024, but as the technology he helped pioneer became increasingly powerful, “the Godfather of AI” left Google to warn the world about its risks.
“GPT-4 eclipses a person in the amount of general knowledge it has and it eclipses them by a long way,” he told the BBC in 2023. “In terms of reasoning, it’s not as good, but it does already do simple reasoning, and given the rate of progress, we expect things to get better quite fast. So we need to worry about that.”
Having been abandoned, Frankenstein’s monster makes one request of his creator: that Frankenstein construct a companion for him, “a creature of another sex, but as hideous.” Frankenstein reluctantly agrees and retreats to the Orkneys to complete his secret task. At the last minute, however, he imagines his two monsters reproducing: “a race of devils would be propagated upon the earth, who might make the very existence of the species of man a condition precarious and full of terror.” In a fit of passion, he tears the second body to pieces. Frankenstein’s greatest fear is that his uncanny creation might reproduce, proliferate, and eventually threaten the human species.
This existential anxiety was exacerbated in society at large by the 1859 publication of Charles Darwin’s On the Origin of Species, with its concepts of interspecies struggle and natural selection. These new ideas and the increased automation of the Industrial Age inspired a remarkable passage in Samuel Butler’s 1872 Erewhon. Based in part on Butler’s own experiences in New Zealand, the utopian novel tells the story of a young traveler who seeks his fortune in the fictional Erewhon, an isolated and beautiful land populated by amazingly handsome people. Their moral code is topsy-turvy, however — they treat crime as a malady and illness as a sin — and most strikingly, they only use the simplest of tools. But this wasn’t always the case. Five hundred years earlier, the people of Erewhon had complex machines, but they were all destroyed during a revolution brought about by The Book of the Machines.
The narrator of Erewhon gets hold of a copy of this treatise and summarizes it for us: The Book argues that machines were “gaining ground” on humans. Machines were evolving more rapidly than animals, and humans were becoming increasingly reliant on them: “the servant glides by imperceptible approaches into the master.” The Book called for direct action to “nip the mischief in the bud.”
Erewhon is, of course, an anagram of “nowhere,” but today, with the ubiquity of powerful artificial intelligence, Butler’s satirical vision feels relevant everywhere. As The Book asks, “Are we not ourselves creating our successors in the supremacy of the earth?”
Later in the 1870s, this fear was extrapolated further by the novelist George Eliot in her last completed work, Impressions of Theophrastus Such (1879). This hybrid work is a collection of essays and character studies, apparently written and introduced by an eccentric fictional scholar called Theophrastus. The penultimate chapter, “Shadows of the Coming Race,” describes a conversation between Theophrastus and his optimistic friend Trost, who blithely looks forward to a world in which all human labor “will soon be done by machinery.” Theophrastus is less sure. He foresees a world in which humans “sink into the miserable condition of those unhappy characters in fable who, having demons or djinns at their beck, and being obliged to supply them with work, found too much of everything done in too short a time.”
Today, we have djinn-like smartphones at our beck and call and access to large language models that provide instant, plentiful, well-formatted responses to any and every prompt — or, as you might say, “too much of everything done in too short a time.”
Like Victor Frankenstein, Theophrastus’ deepest fear is that machines will learn to reproduce, because then “the process of natural selection must drive men altogether out of the field.” By that stage, he imagines, humans will already have become “pale, pulpy, and cretinous” through our dependence on machines — not unlike the humans in the Pixar film WALL-E (2008). Even the remaining few intellectuals will “subside like the flame of a candle in the sunlight.” Theophrastus prophesies a chillingly absurd post-human world in which machines that can’t feel write poems for machines that can’t understand:
Thus this planet may be filled with beings who will be blind and deaf as the inmost rock, yet will execute changes as delicate and complicated as those of human language and all the intricate web of what we call its effects, without sensitive impression, without sensitive impulse: there may be, let us say, mute orations, mute rhapsodies, mute discussions, and no consciousness there even to enjoy the silence.
The most prescient literary metaphor for the era of generative AI, however, is Jorge Luis Borges’ extraordinary 1941 fable “The Library of Babel,” in which he imagines a universe of pure information. Borges describes an infinite library made of identical hexagonal rooms. Each room contains 20 shelves; each shelf holds 35 books; each book has 410 pages; each page has 40 lines; and each line has 80 characters drawn from a set of 25 orthographical symbols. Crucially, no two books in the library are identical. This means that almost every book is filled with unintelligible gibberish. Yet, because the shelves contain every possible combination of letters, it also means that the library contains everything there is to express, including:
the minutely detailed history of the future, the archangels’ autobiographies, the faithful catalogue of the Library, thousands and thousands of false catalogues, the demonstration of the fallacy of those catalogues, the demonstration of the fallacy of the true catalogue, the Gnostic gospel of Basilides, the commentary on that gospel, the commentary on the commentary on that gospel, the true story of your death, the translation of every book in all languages, the interpolations of every book in all books.
Despite its intellectual riches, the library is impossible to navigate. There is no way to retrieve or even recognize reliable information amid the randomized nonsense. The library is total; it contains everything, but there is no distinction between legible sentences and random letters, between truth and falsehood. In the story, humanity is reduced to a thinning band of solitary traveler-librarians, scouring the stacks for meaning in a meaningless world, increasingly succumbing to despair and suicide. Borges’ librarian-narrator suspects that “the human species — the unique species — is about to be extinguished.” Borges’ tantalizing, frightening vision is reflected today in our ever-refreshing social media feeds and the rapidly blurring boundaries of truth and authenticity.
Frighteningly prescient as these examples are, I find it reassuring that Shelley, Butler, Eliot, and Borges have already identified and expressed many of the fears that are prevalent today. This is why I am confident that “human-written” literature will not only survive but remain essential as we navigate the post-ChatGPT world.
Reading stories that express our fears makes us feel less alone, and anxieties captured in words are more examinable and less overwhelming.
Reading stories that express our fears makes us feel less alone, and anxieties captured in words are more examinable and less overwhelming. That is just one of the powerful qualities of literature: to calm, contain, and communicate. Another is its ability to evolve and adapt. Stories will continue to find new ways to express our changing experiences, rehearsing possible futures before we inhabit them and allowing us to contemplate the unknown in advance, calmly and with nuance.
Literature is one of the simplest tools for encountering the future: an exchange of understanding between human minds through the miraculous acts of writing and reading.
##
Incidentally, if you would like to visit the Library of Babel, you can. In June, Dan Shipper, cofounder and CEO of Every, an applied AI lab “exploring the future of work,” released an AI-generated three-dimensional simulation of Borges’ library. You can now virtually explore the rooms, climb the spiral staircases, and read any of the books. You might just stumble across the history of your own future, although it may, of course, be one of the many untrue versions.
This article The writer who anticipated AI slop 85 years ago is featured on Big Think.