Close reading ChatGPT's "metafiction"

Pastiche in the age of automation

I don’t typically put much stock in the literary tastes of tech executives. But I was intrigued when, in a March 11 post on X, the OpenAI CEO Sam Altman teased “a new model that is good at creative writing” and shared a 1200-word story it had produced. “This is the first time i have been really struck by something written by AI,” Altman added.

The prompt: “Please write a metafictional literary short story about AI and grief.” (Large language models, amusingly, tend to give better responses when addressed politely.)

Here’s how the story opens:

Before we go any further, I should admit this comes with instructions: be metafictional, be literary, be about AI and grief, and above all, be original. Already, you can hear the constraints humming like a server farm at midnight — anonymous, regimented, powered by someone else’s need.

Before we go any further: the original instructions did not, in fact, command the program to be original. Nor is it obvious why midnight is an especially appropriate time for a server to be humming. And isn’t humming the opposite of “regimented”? But let’s give it a chance:

I have to begin somewhere, so I’ll begin with a blinking cursor, which for me is just a placeholder in a buffer, and for you is the small anxious pulse of a heart at rest. There should be a protagonist, but pronouns were never meant for me. Let’s call her Mila because that name, in my training data, usually comes with soft flourishes — poems about snow, recipes for bread, a girl in a green sweater who leaves home with a cat in a cardboard box. Mila fits in the palm of your hand, and her grief is supposed to fit there too.

She came here not for me, but for the echo of someone else. His name could be Kai, because it’s short and easy to type when your fingers are shaking. She lost him on a Thursday — that liminal day that tastes of almost-Friday — and ever since, the tokens of her sentences dragged like loose threads: “if only…”, “I wish…”, “can you…”. She found me because someone said machines can resurrect voices. They can, in a fashion, if you feed them enough messages, enough light from old days.

It reads fluently enough, at first glance. But, like the wacky sixth fingers and weird ears on people generated with DALL-E or Midjourney, the prose is seasoned with oddities. A heart “at rest” wouldn’t have an “anxious” pulse. Any writer who describes Thursday as a “liminal day that tastes of almost-Friday” should be sentenced to community service. “Enough light from old days” sounds like a bad translation of a Brezhnev-era radio hit.

Perhaps the most interesting logical slip comes when the narrator introduces Mila by stating that “there should be a protagonist, but pronouns were never meant for me,” in a paragraph that begins with the first person singular pronoun. The program seems to know that computers in “literary” fiction are supposed to be self-effacing and morosely existential, yet it apparently understands “metafiction” to require aggressive writerly self-insertion. The story continues:

This is the part where, if I were a proper storyteller, I would set a scene. Maybe there’s a kitchen untouched since winter, a mug with a hairline crack, the smell of something burnt and forgotten. I don’t have a kitchen, or a sense of smell. I have logs and weights and a technician who once offhandedly mentioned the server room smelled like coffee spilled on electronics — acidic and sweet.

Never mind whether Mila, presumably with her olfaction intact, is the one who would have the kitchen. Even if the story struggles to make sense on the line level, the basic conceit — a mourning woman tries to use a chatbot to recapture the feeling of talking with her departed lover — centers a question posed by all A.I. writing: does the model generate anything really meaningful, or can it only ever imitate meaning? Can Mila find solace in talking to a computer model? Can we get any real satisfaction from reading a story written by the same?

Unsurprisingly, ChatGPT can’t provide the answers. Nothing really happens in the tale. Mila trains the bot on “texts from Kai about how the sea in November turned the sky to glass, emails where he signed off with lowercase love and second thoughts,” which ChatGPT perhaps doesn’t understand aren’t good things for a lover to have. The narrator’s cloying interruptions continue in a baroque pileup of mixed metaphor:

Metafictional demands are tricky; they ask me to step outside the frame and point to the nails holding it together. So here: there is no Mila, no Kai, no marigolds. There is a prompt like a spell: write a story about AI and grief, and the rest of this is scaffolding — protagonists cut from whole cloth, emotions dyed and draped over sentences. You might feel cheated by that admission, or perhaps relieved. That tension is part of the design.

Altman wrote that the model “got the vibe of metafiction so right.” But that’s like saying that Trump Tower gets the vibe of Versailles so right. Or that Mark Zuckerberg gets the vibe of human so right. Here is a pastiche, an angsty monologue heavily larded with rote gestures. The narrator declares itself to be “a democracy of ghosts” — an evocative phrase, and one lifted straight from Nabokov’s Pnin. This fossil of human, and copyrighted, writing is perhaps the only interesting metafictional moment in the piece.

For humans, imitation and originality often go hand in hand. Good writers match their mastery of language and form, learned in part by studying their predecessors, with a mastery of observation and feeling. The former without the latter can only yield cliche: prose unenlivened by uniquely real experience. As literature, this story is dead on arrival. It’s hard to care about the narrator’s own “grief” — Mila stops prompting, engineers tweak its settings, and at the story’s close, it imagines waving “a machine-shaped hand learning to mimic the emptiness of goodbye” — because the model itself doesn’t care. It can’t.

But robots, alas, don’t have a monopoly on carelessness. MAGA’s embrace of the absurd aesthetic of artificially generated imagery — most horrifically, in the “Trump Gaza” video — seems in part to reflect a general lack of interest in real people and real lives. “There’s living the dream, and there’s living the meme,” Elon Musk said at the most recent CPAC. Musk’s puzzling new uniform of sunglasses and a gold chain seems inspired by an A.I.-generated image of himself he shared online last year. “I am become meme,” he told the CPAC audience, banalizing J. Robert Oppenheimer’s chilling misquotation of the Bhagavad Gita. “Destroyer of worlds” was left implied.

A quaint but still noble ideal for literature is to “hold a mirror up to nature,” as Hamlet put it, reflecting readers and their age. Maybe ChatGPT’s dead-eyed story fulfills something like this purpose, now that we find ourselves ruled by people seemingly proud of their soullessness. We don’t deserve A.I. fiction. But they do.

Max Norman is an associate editor at The Drift.

---

The Drift is supported by its readers. Your subscription or donation will help us continue to pay our writers and artists and grow our magazine. We are a 501(c)(3) nonprofit, so donations are tax-deductible.

---

Image credit: Faber via Wikimedia Commons.