For about two minutes last week, I accidentally found myself in the company of Margaret Atwood, Stephen King, Salman Rushdie, Han Kang and other stars of the literary world. A data set of about 183,000 books by identified authors had been used “without permission to train generative-AI systems by Meta, Bloomberg, and others”, The Atlantic’s Alex Reisner reported.
Many people knew that this collection, dubbed Books3, has been used for decades to train AI language models, along with other data sets, but Reisner made it possible for us to search and see what it actually contains. I found books by countless writers I admire, from Elena Ferrante to Neil Gaiman, Perumal Murugan and RO Kwon — and one of my own, a fantasy fiction featuring Delhi’s stray cats, dogs and bandicoots, The Hundred Names of Darkness (2013).
Writers have an edge in the war of words, and outrage flowed freely. The novelist Lauren Groff wrote on X, the platform formerly known as Twitter, “I would never have consented for Meta to train AI on any of my books, let alone five of them. Hyperventilating.” Margaret Atwood, responding to the fact that 30 of her books were lifted, tweeted: “That is so . . . cheap.” The Booker-winner Richard Flanagan felt, he wrote, as though “my soul had been strip mined and I was powerless to stop it”.
My own dismay and rage have lasted. The casual way in which all of our works were taken without permission and tossed into the AI blender feels like a violation. The one novel I have in this mix is hardly War and Peace — but it stings that none of these corporate behemoths thought to ask thousands of writers if we wished to contribute to AI models that might eventually replace us.
This month, 17 of the world’s most famous fiction writers, including David Baldacci, Jodi Picoult, John Grisham, Rachel Vail, George RR Martin, Jonathan Franzen and George Saunders, joined a lawsuit filed by the Authors Guild against OpenAI. “Writers should be fairly compensated for their work. Fair compensation means that a person’s work is valued, plain or simple,” Saunders wrote in a statement.
Writers have now joined artists, photographers, film-makers and musicians to battle the encroachment of generative AI. Most creative professionals have an intensely personal relationship with their work; you write from the depths of your imagination, the far shores of your experience. But for many in the tech world, writing, like art and music, is apparently free content to feed into the machine.
This July, Sam Altman, OpenAI’s CEO, tweeted, “Everything ‘creative’ is a remix of things that happened in the past, plus epsilon and times the quality of the feedback loop and the number of iterations.” It’s a view of creativity as a commodity that can be extracted from humans and remixed by machines. In a 2016 interview with the New Yorker, Altman speculated that “computers will have their own desires and goal systems”, and that intelligence can be “simulated”, having already concluded that the idea of human uniqueness is obsolete.
A tiny indicator of the value Silicon Valley places on writers is apparent from a report in the online tech magazine Rest of World. Last month, it was reported that two data giants, Scale AI and Appen, ran ads asking for poets and writers to write original stories to train AI language models. Rates were as low as $1.43 an hour.
Last week, many celebrated as the Writers Guild of America ended one of the longest strikes in its history. In a major concession, Hollywood studios have agreed that while writers may use AI if they wish, “AI can’t write or rewrite literary material, and AI-generated material will not be considered source material”. But generative AI is like the invention of the internet — transformative and surely unstoppable. This genie isn’t going back in the bottle. And of course the idea of generative AI is alluring — the promise is that, by feeding a few prompts into the belly of the beast, anyone will be able to write like Chekhov, make art like Frida Kahlo, sing like Aretha Franklin.
As a young reader, I was hopelessly in love with books, dreaming of writing my own some day — not for fame, but for fun. Three novels later, I know that what I truly love is writing itself: the messiness of it, the pleasure of craft, the failed drafts, the characters who wander into your imagination and live with you for years.
Generative AI might be able to digest a million books and produce a million more. But only a tech billionaire would want to invent a machine that replaces the best part of being a writer: finding the right words, yourself.
Join our online book group on Facebook at FT Books Café