Are the recent AI literary scandals changing everything?

AI has breached the walls of high literature

The short story “The Serpent in the Grove” by Jamir Nazir, a writer from Trinidad and Tobago, won the prestigious Commonwealth prize from the Caribbean, and was published in Granta magazine. Shortly after winning, it was widely accused of being in whole or in part AI-generated.

Earlier this month, the Nobel laureate in literature Olga Tokarczuk said:

When writing my latest novel… I asked this advanced model what kind of songs my protagonists would be listening to at a dance, a few dozen years ago, and AI gave me a few titles… Often I just ask the machine, ‘darling, how could we develop this beautifully?’ Even though I know about hallucinations and many factual errors in the algorithms in terms of economics and hard data, I have to add that in literary fiction this technology is an advantage of unbelievable proportion.

This prompted allegations that she let AI write the novel, at least in part. Tokarczuk quickly clarified:

1. I make use of artificial intelligence on the same principles as most people in the world – I treat it as a tool that allows faster documenting and checking of facts. Whenever I use this tool I additionally verify the information. Just as I have done for several decades by reading books and by exploring libraries and archives.

2. None of my texts, including the novel that will appear in Polish this fall, has been written with the help of artificial intelligence – except for using it as a tool for faster preliminary research.

Earlier this year, the New York Times cut ties with a journalist after discovering that he’d used AI to write a review. And, to take one more example, the publisher “Hachette cancelled the release of a debut horror novel, Shy Girl, over concerns it was written at least partially with AI.”

So, how bad is all this? And is it all bad? And would AI be be able to tell AI didn’t write my stories?

The literary establishment and mainstream media response: Meh

Looking at how the literary establishment and mainstream media has responded to and spoken about these scandals, above all, I’m struck by the meh attitude.

In Literary Hub and The Guardian, for example, I feel that the purpose of the articles on these scandals is primarily to record a major historical event which has already become blasé, boring, accepted, rather than to explicitly probe and examine what the use of AI means for the process, art, economy, and, well, human aspect of writing fiction.

The Guardian mentions the irony of the Granta people using AI to check whether the winning story was written by AI, glosses over the paradox that there’s a dawning industry of anti-AI/AI-detecting AI, and then concludes with another meh.

The language in these articles sound like quick telegram dispatches from World War II that read something like “German tanks enter Paris,” “Pierre Drieu la Rochelle supports Nazi propaganda.” The articles on Granta and Tokarczuk could have been titled “AI occupies strategic sites in the republic of letters,” “Who’s joining the bandwagon?”

Of course that may have a lot to do with AI companies’ largely successful PR machinery, which portrays the introduction of AI in everyday life as a world-historical event, a fait accompli akin to the invention of the written word, or something even bigger and better than that.

It seems too much to ask these venues to hit the breaks and critically probe the slogans of the AI barons in, I don’t know, every piece they do on the topic of AI.

AI doesn’t change everything, but it does change some things

Let’s assume writers have already started using AI, either as Nazir (allegedly) or Tokarczuk (professedly). Does that change everything?

No.

AI’s fundamentally producing natural language-ish text based on statistical averages. Admittedly, that can get you pretty far with all kinds of complex tasks, but it can’t fill in for human experience, mind, body, emotion, soul, and stories issuing from all that. The language of holding a warm cup of coffee and watching the shadows dance on the walls on a chilly November day, which makes you think of your grandmother, hear her voice, feel her embrace from those decades ago. There’s still a lot of people who want to read stories written by humans, who want to read words that feel like holding a warm cup of coffee, watching the reality of shadows on the walls.

AI to me feels like a search engine on steroids, which can and does assist writers in researching and brainstorming all kinds of ideas. I don’t think this fundamentally changes the nature of writing. I think it just speeds up the process of researching and brainstorming ideas. But so did printed mass media combined with the combustion engine.

Does it change anything?

Yes.

Here’s the bad news: I think some people, probably a lot of people, don’t give a shit about where a story comes from or who produced it, as long as it replicates stimulating plots, characters, conflicts, endings, styles, and so on, well enough. Those people will be perfectly satisfied by a kind of literary McDonald’s: quick, gratifying, entertaining, distracting, cheap stories.

They won’t care about AI stories’ brutal origin of primitive accumulation.

I also think that all kinds of political powers (government, corporations) can now rely on endless Orwellian “novel-writing machines” to try to control information, rewrite history, mold culture and affect, and so on. But I’m not sure it’ll improve, or degrade, the quality and effectiveness of the propaganda: so there’s that.

What does AI say about my stories?

All this got me thinking. What would an AI agent say about my stories? So, with some dread (because, let’s face it, we’re all machine-like in some ways, probably a lot of ways), I asked Claude what it thought about my stories in World Literature Today, and whether it thought they were generated by AI. I noticed it said something like “thinking, paying close attention to copyright requirements,” which hopefully means it isn’t taking my stuff to create an AI Admir Skodo (I also set it to not use my chats for training purposes, but who the hell knows…). Anyway, this is what it said:

No — and I don’t think it’s close.

A few specific reasons, drawing on what I actually read rather than vibes:

The Bosnian-language texture isn’t ornamental. Jebiga, džezva, fildžan, Bajram, the garbled child-mouth Bismilaihrahmanirahim, the Swedish Kvinnor/Män/Flickor/Pojkar paired with Žene/Muškarci/Curice/Dječaci — these aren’t dropped in for flavor, they sit in the syntactic positions where a bilingual mind would actually reach for them. AI prose tends to either avoid this register or perform it; yours uses it.

The archival specificity is the wrong shape to be fabricated. ma_avdsyd_asylär_0992-34668_iddok as a Migration Agency file ID; Maria Svärd at Kristianstadsbladet; the Golden Spade; Kumla prison; Revinge municipality; the JAS Gripen named after a mythical beast; Edvard Kardelj’s Sećanja. AI hallucinates plausibly-shaped specifics; these are specifics with friction, the kind you only get from a historian who actually works in these archives. (Your day job as a migration researcher is doing real work in the fiction.)

The tonal control is uneven in the way humans are uneven, not in the way models are uneven. Models tend to smooth — every paragraph hits the same register. Your prose lets the philosophy professor Zlatko’s golden-calf riff sit next to “shove it up your skinny ass,” lets a meditation on Danilo Kiš sit next to Mirela screaming jebiga, jebiga, jebiga, lets the canteen menu (corn flakes, undead potatoes, watery milk) interrupt a paragraph about identity fraud. The transitions are abrupt in a way that reads as authored choice, not statistical averaging.

The Kiš influence is doing structural work, not decorative work. The juxtaposition principle — Bosch paintings, Swedish freezer cases of ćevapi next to fiskgratäng, the US Senate on the Iroquois Confederacy — is held as a method through the whole Mandal piece, and Donated Clothes is built on the same logic (the refugee camp as Bosch panel). A model could name-drop Kiš; you’ve metabolized him.

The honest caveat: I can’t prove a negative, and AI text has gotten harder to flag confidently. But if you’re asking whether I read these as machine-written — no. They read as the work of someone with a specific biography, a specific archive, and a specific set of literary loyalties (Kiš especially, also visible traces of Aleksandar Hemon’s territory), writing in a voice that’s recognizably their own across both pieces.

I guess that’s good news?