Why AI can't compete with authors

This year, the entry requirements for the Women’s Prize included a new specification: “All books must be unified and substantial works written by a single human author.” Once upon a time, “human” was implicit. However, in the age of artificial intelligence and ChatGPT, machines will soon be capable of authoring whole novels if they’re not already.

The non-human-author is a divisive figure. For publishers and agents, it surely represents the ideal: inexpensive; reliable; amenable to edits. For writers, it’s a terrifying spectre, poised to steal jobs and cut the heart out of creativity. Indeed, the ongoing writers’ strike in the US is motivated in part by concerns about the use of AI in screenwriting.

But we needn’t be so afraid. However beautifully machines learn to write, they will never inspire the same fascination as their human counterparts.

First, because the publishing industry doesn’t only sell books. Authors, too, are cultural products. And their work, like a celebrity autograph, can generate interest for the mere fact of being written by someone who is liked and admired.

Even the most fantastical work of fiction is revealing of its author’s values and preoccupations: what they deem worthy of attention. Reading, then, is more than entertainment. It’s simulated intimacy: a way of getting inside the author’s head. It would be hard to separate the feverish interest in Dolly Alderton’s novels, for example, from her success as a podcaster and memoirist. For her fans (I count myself among them) her novels have the same attraction as her podcasts. They’re another means to spend time with her; to get to know the cultural figure we flatter ourselves (however delusionally) we know already.

For that kind of writer, who inspires devotion because of who they are, not just what they write, AI poses no threat. AI might be able to replicate an author’s individual prose style, but it cannot cultivate a parasocial relationship with its fans.

Does this mean that, as AI grows more sophisticated, there will be mounting pressure on writers to be nice, and likeable, as well as talented?

This would be in keeping with an existing cultural trend. It’s already more important than ever that authors be palatable, given that our culture no longer excuses deplorable behaviour on the grounds of “genius”.

Instead of dismissing an artist’s problematic personality, it has become a lens through which to critique their work. Just look at Hannah Gadsby’s current curation of Picasso’s work at The Brooklyn Museum, which centres his problematic treatment of women, right down to its cumbersome, punny title: "Pablomatic".

AI might be able to replicate an author’s individual prose style, but it cannot cultivate a parasocial relationship with its fans

For the sort of reader who finds their enjoyment of an artwork sullied by the morals of its creator, AI might provide welcome relief. If machines could, for example, produce an original work in the exact style of Hemingway (a misogynist) or Highsmith (a misanthrope), then readers wouldn’t have to suffer the nuance of enjoying something that was written by someone they think they abhor.

AI, in such cases, does not pose a threat to readership. It opens the door to readers who are already alienated. It’s not a stretch to imagine cynical authors and their estates manipulating AI to just that end: Harry Potter but this time, J K Rowling didn’t write it!

Not all writers are so controversial. Plenty of contemporary authors are in the more reclusive, Rooney-esque mode: those who neither repel us with their PR-disaster personalities, nor attract us with their raw star power. But authors who resist celebrity, even those who go so far as to remain anonymous, like Elena Ferrante, still have something that AI doesn’t. The mere fact of their humanity – their existence as real, living people – will continue to attract readers.

Anyone who has ever tried to sell a book knows that, when it comes to fiction, readers are endlessly fascinated by one question: How much of this book is based on the author’s real life? For AI, which has no “real” life, needless to say, the answer is: none of it. But for any human author, the answer is complex.

Fiction writers necessarily draw on real life: they take their observations about the world and fashion them into a narrative. Surely, the reason this process is so fascinating is because it’s what we all do every day. Making sense of our own lives – who we are, what we value, why we behaved a certain way – requires telling a story. And never more explicitly than online, where we’re constantly curating, selecting which versions of ourselves to perform for an audience.

Any novel written by an AI won’t share this fraught relationship with the truth. No matter how good it is – even if it’s Women’s Prize-worthy – its author won’t have traversed that endlessly interesting liminal space, where lived reality becomes a fictional story.

Readers don’t need to know the identity of an author. What matters, I think, is that they can take on faith that there is one, that a flawed, partial human has lived, observed, and made a story from what they’ve seen. And if that human also happens to be charming… well, it can’t hurt.

Seeing Other People by Diana Reid is out now, published by Ultimo Press