Stray Thoughts on a Writing Contest Likely Won by A.I.

I’d clocked the story of the Commonwealth Foundation awarding a prize to a short story that was very probably written by A.I. It was sometime in late May or early June. After that, I heard it mentioned on various bookish podcasts with a resigned and sad welp-it’s begun tone, as if this is our future going forward and there’s very little we can do about it. (Which seems true enough, for what it’s worth.)

More recently, I was reminded of the story by a piece in Slate that basically just made fun of the writer’s half-assed attempts to explain away the controversy.

A Writer Was Accused of Cheating to Win a Prestigious Award. His Defense Has Only Made the Case Stranger.

Jamir Nazir pulled out all the stops in an attempt to dispel accusations that A.I. wrote his Commonwealth Prize–winning story “The Serpent in the Grove.”

SlateLaura Miller

Once again engaged in the drama, I went looking for even more down this particular rabbit hole, and I found this wonderful piece in The New Yorker by Katy Waldman.

Did a Chatbot Write a Prize-Winning Story? Does It Matter?

A Commonwealth Prize-winning short story appears to have been generated with A.I. The response exposes larger fears about honesty, motivation, and expectation in literary writing.

The New YorkerKaty Waldman

At the very beginning of her article, Waldman uses one of the most widely circulated of the head-scratcher lines from the story, “The Serpent in the Grove,” by Jamir Nazir of Trinidad:

“They called her Zoongie,” one passage from the story goes. “Maybe it was a name; maybe rain took a shape and decided to keep it.”

Another hum-dinger is revealed in Waldman’s high-level summary of the prize-winning story’s plot:

“The Serpent in the Grove” is about a farmer, Vishnu, who tricks his wife, Sita, into falling down a well. The story has its share of glaringly nonsensical phrases that should have tipped off anyone paying an iota of attention—for instance, when Vishnu spies a sexy visitor, we learn that the woman “had the kind of walking that made benches become men.”

When I read both of those examples, I can’t help but think about National Treasure. In one of the early scenes, after hearing what he perceives to be a bunch of nonsense, Ben Gates’ lovable sidekick, Riley, blurts out: “Albuquerque. See? I can do it, too. Snorkel.”

It became my and my brother’s shorthand for responding to nonsense with nonsense, perfectly applicable any time a person is just spouting rubbish.

Zoongie put me on alert for it, but I don’t know the ins and outs of every culture’s names, so I held my tongue. But then the text itself calls into question whether Zoongie is a legitimate name, and seems to come down on the side of “Who can say?” For, we learn, “maybe rain took a shape and decided to keep it.”

Let’s think hard about what it’s saying. At first, I though the “rain took a shape” bit referred to the human known as Zoongie. As in, raindrops become human-shaped. But if you look at the sentence, it seems more and more like “rain took a shape” refers to the word “Zoongie” — “Maybe it was a name.”

Maybe I’m overthinking it. Let’s look at the next one: a woman “had the kind of walking that made benches become men.”

So, clearly the walking is eye-catching, alluring, sexy even. But benches? What is the connection there? What is the imagery we’re supposed to see in our mind’s eye? My most charitable reading is that it was seductive enough to bring something inanimate to life, maybe? Her essence was enough to breathe life into a random object? That’s the best I can do for it. Otherwise, snorkel.

The cautious side of me — the but-what-if-we’re-all-wrong? side — found Waldman reading my mind when I got to this paragraph:

Maybe the most fascinating aspect of the story—fascinating because it feels inhuman in a plausible and almost endearing way—is what seems like a fundamental confusion about the kinds of behavior, agency, or interiority that one might expect from inanimate objects, body parts, or concepts. “The grove remembered,” we’re told. “Water is jealous.” “Wood complained.” Online, commenters mocked the description of Vishnu’s fellow-villager, Marsha, who is “big in the way of women who never apologize to furniture.” They lampooned the sentence: “Hard living lays itself on a man like wet sacking; it never asks permission.” Yes, these lines are inelegant, but part of what makes them so strange is what they presume about the inner lives of things and abstractions. How, after all, would “hard living” ask permission? Why would you apologize to furniture?

As in, maybe, just maybe, isn’t it batshit crazy enough to be real? Who wouldn’t slap the side of their computer monitor if their L.L.M. of choice spit out something like these examples? (Of course, if it were used, surely it was prompted to write something full of literary-sounding imagery and symbolism; the average user saying “Draft an email for me” isn’t likely to get such off-the-wall efforts.) I do find it interesting that one of the things that we use for “proving” a piece of writing is real is a sense of unpredictability (predictability being the lifeblood of L.L.M.s); something unexpected, something specific, something out of the ordinary, something that hasn’t been said a million times over. Waldman seems to be with me on this one: What if it’s not so wrong because A.I. got it wrong, but rather a human, imperfect as we are, got it wrong?

If we’re to give Nazir and others who come under scrutiny this kind of grace, then our bigger issue is this: It’s just not very good writing. Full stop.

If A.I. wrote it, then we rail against it as bad and lazy. But if Nazir wrote it, we would be just as entitled to rail against it as bad and intentional. The world was run amok with bad-but-intentional writing in the blessed beforetime when no such thing as an L.L.M. existed. Bad-but-intentional is not the same thing as “intentionally bad”; bad-but-intentional is the very real condition of wanting to be better at writing than one actually is.

Stephen King, in On Writing addressed this very topic, lest he be accused of slinging snake oil that would help anyone who bought his book to write and sell as many as he has:

There are no bad dogs, according to the title of a popular training manual, but don’t tell that to a parent of a child mauled by a pit bull or a Rottweiler; he or she is apt to bust your beak for you. And no matter how much I want to encourage the man or woman trying for the first time to write seriously, I can’t lie and say there are no bad writers. Sorry, but there are lots of bad writers.

Waldman points out an even scarier possibility: There are lots of bad readers, too. In fact, some of them seem to be judging prestigious literary contests:

For many commenters, these rhapsodically expressed banalities are less offensive than the fact that a group of cultural gatekeepers rubber-stamped the story—that the serpent got into the grove in the first place. As the novelist Will Self and others have written, Nazir’s success suggests an unhealthy literary culture, one that was deteriorating long before the A.I. asteroid hit (and maybe since the dawn of literary culture). Some X users said that the winning entry embodies the tendency of M.F.A. programs to promote a kind of stylistic polish at the expense of substance. One of the judges, Sharma Taylor, praised Nazir’s story in terms that, to many readers, sounded suspiciously bot-like, remarking on his “precise yet richly evocative” language, which conveys “vivid, lush imagery with remarkable economy.” That Taylor’s statement also conforms to the conventions of creative-writing-seminar plaudits is part of the problem.

I don’t know exactly what this all means or exactly how we got here, but I can say with absolute certainty: None of this is good for literature.