It’s that time of year again. The giddy middle of May. When millions across the English-speaking world gather to find out who has been shortlisted for the Commonwealth Short Story Prize.
This year’s shortlist, drawn – as ever – from a diverse selection of not-European, not-male authors, is particularly enriching and profound. As the committee itself puts it, the stories ‘bring compelling characters to life in sharply drawn settings, exploring themes of power, family tension, resistance and unheard voices, alongside courage and unexpected connection. Among them are a keenly observant domestic worker, a young woman whose henna art enables silenced women to speak, and a resourceful young sheep farmer’. All of which makes it especially surprising to discover that at least one of the stories was allegedly written by a data centre in Texas.
The story in question is ‘Serpent in the Grove’. The author is Indo-Trinidadian, and his biography on the Granta website – which has published the piece as the winner of the Caribbean regional award – describes him as ‘a prolific poet and author, with books published and others forthcoming, he is particularly known for his love of poetry.’ I’ll spare you the name. He is being eviscerated on social media, and I genuinely feel sorry for him.
The allegations rest on several pieces of evidence. The weakest is that AI detectors flag the prose as 100 per cent machine-made. As someone who has used and written about these tools for years, I can tell you the detectors are highly unreliable. Their failure rate is significant: they regularly mark fully human writing as wholly artificial, and vice versa.
More persuasive is that the author has not previously shown great literary skill – he is a self-published author of unremarkable love poetry. Meanwhile, a few months ago, he began suddenly enthusing about AI, mainly on LinkedIn. Something of a coincidence.
The better evidence, though, is the story itself. AI, used incompetently and prompted badly, produces a particularly irritating style: deeply florid prose overstuffed with similes, tell-tale constructions, and metaphors that sound passable or even pretty at first glance – but fall apart on scrutiny.
Here are a few choice examples from ‘Serpent in the Grove’, so you can judge for yourself. It opens: ‘They say the grove still hums at noon,’ and it goes on in that style. The sun ‘beats until the roof talks back in a dry moan.’ Coffee grows on a slope that ‘wanted rain in teeth.’ The heroine has a walk that ‘made benches become men.’ Her friend is ‘big in the way of women who never apologise to furniture.’
There is more, much more. Some lines make no sense at all. ‘Water is jealous.’ ‘Doing is a treacherous bridge.’ ‘First good rain after dry is a forgiveness the sky gives itself.’ ‘She didn’t hate her leg. It had thrashed exactly long enough to catch a stone.’
Is this definitely AI? Without a confession from the author, Granta or the judges – or sight of the author’s prompts, if they exist – we simply cannot know. Neither the author, the judges nor Granta have responded to these allegations at the time of writing. But in a sense, it doesn’t truly matter. Even if this particular story isn’t AI, at some point AI will win a literary competition. And the next time, we genuinely won’t know.
This is not the first occasion I’ve made this argument. In January 2023 I published a piece in The Spectator called ‘AI is the End of Writing’ which earned me much grief. I was scorned and scolded, and told AI could only produce rubbish, and would never match a proper human writer.
AI, used incompetently and prompted badly, produces a particularly irritating style
My critics were half right. AI back then was mostly producing rubbish, unless you got very lucky. But my critics also made the mistake people always make about AI: they forgot that AI would get better, and better. They failed to extrapolate.
Here is how I put it back in 2023:
The machines will come for much academic work first – essays, PhDs, boring scholarly texts (unsurprisingly, it can churn these out right now). Fanfic is instantly doomed, as is self-published fiction. Next will be low-level journalism, copywriting, marketing, legalese, tech writing; then high-level journalism will go, along with genre fiction, history, biography, screenplays, TV drama, drama, until eventually a computer will write something like Ulysses, only better. The only prompt will be: “write a long amazing novel on whatever.”
Three years on, I stand by every word. If anything, it’s all moving faster than I predicted. One of the ironies of ‘Serpent in the Grove’ is that, if it is AI, it is unnecessarily bad AI. These days, if you prompt the right models carefully and precisely, you can now produce genuinely good prose, sometimes startlingly good.
As for now, however, you still need a human in the mix – and not just any human, but a talented one. We are entering literature’s version of ‘centaur chess’: the era when a great chess player working with a computer beat any human or any computer playing chess alone. In the coming years, a good writer using AI will outperform both an unaided AI and an unaided human.
And yet, that era will also end. Literature will again follow chess. The AIs will get better, then vastly better, then perhaps so much better we won’t grasp how much better they are – the way most people don’t really like Ulysses, or even get it, because it’s too richly complex.
All of which sounds like a very depressing note on which to conclude. But consider the cheerier reading. The Commonwealth Prize exists to amplify unheard voices. And what voice in human history has been more thoroughly unheard than a large language model whirring 24/7 in a vast air-conditioned factory near San Antonio? The judges may have been more progressive than they realise.
Comments