A prize for AI?
Last May, the literary world was rocked by the news that the winner of the Commonwealth Literary Prize—a short story titled “The Serpent in the Grove” by a little-known writer from Trinidad named Jamir Nazir—had been written using AI. Words like “fraud” and “hoax” flew out quickly into cyberspace. Probably suspecting something was off, someone had used an AI-detector app called Pangram to scan the piece, which returned an unequivocally damning judgment: “100% AI-generated.”
Along with other Commonwealth winners, the story had been published in the respected literary magazine Granta. Given the uproar, the Commonwealth Prize judges reviewed Nazir’s drafts, and pronounced the work authentic; Granta severed its ties with the foundation giving the prize, but has retained the story on its website anyway for all to see and evaluate.
And that’s exactly what I did—go to the website and read the work for myself before coming to any conclusion of my own.
Now, it’s not as if literary hoaxes are anything new; it’s happened before, and it’s even happened here. Long before AI, plagiarism and outright invention (a crime in the case of nonfiction) were practiced not only by the lazy and untalented but also by the cunning. Among the most notorious cases in the US were those of James Frey, who fabricated large sections of his bestselling memoirs, and Janet Cooke, whose Pulitzer-prizewinning story for the Washington Post turned out to be based on an entirely fictional character. Here in the Philippines, a postal clerk duped generations of Filipinos into believing that the Code of Kalantiaw and a novel by Fr. Jose Burgos titled La Loba Negra were the real thing.
The fraud in these cases was uncovered not by programs like Pangram or GPTZero but by dogged scholarship—in the case of Kalantiaw, through the work of the historian William Henry Scott, who showed that Marco’s “code” had words that weren’t in use until centuries later.
Even with software, proving literary deception today is a lot more difficult, because creative writing almost by definition is, well, deception of the highest order, extremely liberal in its methods and means, and devious (without being necessarily evil) in its motivations. Fiction especially is intrinsically ambiguous, unlike journalism which rests on fact. So how can anyone be sure that this prizewinner was faked?
“The Serpent in the Garden” is not a very long short story at less than 3,400 words, and the controversy aside, is certainly an interesting if not significant one in many ways: freshness of language, simplicity of plot but complexity of character, power of imagination. It talks of a man named Vishnu who, enamored of a vamp he meets at the bar, contrives to drown his wife Sita in a well; she falls into it but then manages to crawl out with a neighbor’s help. Sita and even their young son seem to understand exactly what has happened, but no one brings it up and they somehow manage to survive into old age keeping that terrible truth unspoken.
So would I have given the story a prize? That would depend on its competitors, which I haven’t seen, but it’s surely a story I would teach, both for its own merits and for its alleged employment of AI. I might even do that this incoming semester.
Some people have savaged the story after the scandal broke, in the way people change their minds instantly about someone after being told that person served time for rape or murder. Personally—and this could well be an unpopular opinion—I found it rich and engrossing, at least partly because of very same startling metaphors that got flagged as AI giveaways. It’s not an easy read, with its localized English, and I could see how Western readers attuned to the kind of language you hear at Walmart would find it painfully artificial (my friends in the Midwest found Nick Joaquin “too lush” when I had them read him). I don’t write that way myself, which is probably why it intrigues me even more. Here are some random excerpts:
“Sita moved quiet as if sound were taxed. Nineteen and brown like dust after rain, she turned roti dough with a rhythm that came not from joy but from endurance.”
“Marsha lived two bends down. If the village had a mouth, it was hers. Big in the way of women who never apologize to furniture, she had a laugh that shook dust from joists and a voice that could soften to coax a child from a ledge. She knew the ways of men hollowed by want until only one thing remained. She noticed the fresh-cut path and the way land bore witness. People talk about bush like it dumb. But bush keeps memory the way hair keeps scent.”
“Sita lifted two planks and slid them aside. Wood complained in a voice too near speech. She lowered the pail until rope slackened. Smell rose —old wet, crushed jasmine, frog skin. On the second haul, the board beneath her shifted the way a tired man shifts in his sleep. The plank gave one long groan and swallowed its word. Stone, shoulder, hip; shock of cold tearing breath. One foot banged and screamed. The wall was slick as lizard. She clawed moss and slid. Water took her and would not return her.”
So what did the software flag? Unlikely metaphors, it said, like the vamp “had the kind of walking that made benches become men.” Or that a character “smiled like sunrise over a sink.” It noted that there were nearly a hundred metaphors in such a short piece. It claimed that in the sentence “Coffee and cocoa leaned wild on a slope that wanted either rain in teeth or none at all,” the phrase “rain in teeth” made no sense, was a hallucination.
But at this point, my own gut was telling me to trust Pangram less than the author. Call me dumb, but I find “rain in teeth” rather fetching. Any first-time workshopper would catch that.
So would I have given the story a prize? That would depend on its competitors, which I haven’t seen, but it’s surely a story I would teach, both for its own merits and for its alleged employment of AI. I might even do that this incoming semester, when, for the first time in many years and possibly the last, I’ll be teaching an undergraduate class in Professional Writing. It’s a course I designed and had been teaching for two decades, but the big difference this time is that we’ll be dealing with AI—which is something that I don’t intend to ban outright (like Canute commanding the waves) but to grapple with head-on, negotiating the line between one’s imagination and what AI might do. I can foresee how controversial this will be, and I recognize the adamant opposition to AI from many writers and artists on both aesthetic and ethical grounds, but at 72, this could be my last look over the horizon. (I’ve already declared that I will no longer judge literary contests nor take part in literary workshops except in my own classes, so this will be my arena for engaging with AI.)
I wouldn’t be surprised if, in the future (and if it hasn’t happened already), some smartass with an ax to grind enters a wholly AI-generated story or poem into the Palancas, only so that when trusting judges make it win, he or she can declare them and the entire literary Establishment (yes, that cabal of superannuated gatekeepers who keep ignoring one’s literary genius) a bunch of fools.
Right now, what I’m thinking is, if you can come up with the kind of prompt that produces “The Serpent in the Garden,” then you deserve a prize —maybe not the Commonwealth one, which remains an open question, but at least one for inventiveness and audacity which seem to remain in short supply, with or without AI.
