Did a Human Write This Column?

‘I’m reading a book about anti-gravity. It’s impossible to put down.”

“I used to be a banker, but I lost interest.”

“I told my doctor I broke my arm in two places. He told me to stop going to those places.”

If you are a red-blooded American, you rewarded these dad jokes with at least a modest chortle. But are they less funny if I told you they were written by Claude, America’s most fashionable new artificial intelligence program?

Part of what makes a dad joke funny is the process by which it is midwifed into the world. The joke is created in a brain, slips past a cranial censor, and out of a mouth. The funniest part is knowing that a specimen of Homo sapiens was actually proud of a corny one-liner. And it’s hilarious that dads try to embarrass their children by firing off so-so wordplay as if it were worthy of an HBO comedy special.

But Claude has no brain. It has no children to horrify. It doesn’t stand near a grill, wearing shorts with dark socks and Birkenstocks, lecturing people on the proper way to sear meat. So what is a dad joke without a dad? Or any other human being?

The meaning we should impute to text created by AI is now the primary debate in the publishing world. In the past week, Hachette Book Group pulled the forthcoming horror novel Shy Girl after allegations that the author, Mia Ballard, had used AI to write portions of the text. The New York Times fired book reviewer Alex Preston after realizing that his review of Jean-Baptiste Andrea’s Watching Over Her closely resembled a review that was published in another paper — and that AI had repurposed. (Conversely, given the rancid reviews of her most recent book, “Millennial feminist” Lindy West might want to claim that her most recent memoir was written by AI just to convince people she’s not that bad of a writer.)

A strong consensus is forming: Being caught using AI is a death sentence for authors. The purpose of being a writer, after all, is to convey to the reader one’s own thoughts, feelings, and opinions, not the impersonal stylings of a computer language model. For a published writer to borrow passages from AI is akin to a bride and groom using ChatGPT to write their wedding vows. (People actually do this.)

But there are AI defenders. When the Washington Post’s Megan McArdle said that she used AI to help her do research for her columns, social media users flew into a rage. (Many of the accounts castigating her were, no doubt, AI chatbots.)

Yet McArdle says no chatbot ever touches her copy, which seems fair enough. If you use Google to find links or do research, you are using some form of artificial intelligence. Some AI text editors can be handy in helping lower-level writers structure sentences and spell correctly. (I have, however, expressed serious worries about the long-term effect of these AI editors on the English language.)

The Toronto-based writer Stephen Marche, who bragged about using AI to write text within his novel, wrote (we think!) a piece for The Guardian in which he shouted his reliance on algorithms. He claims he wrote his book using prompts but that the process was “no easier than simply writing.”

Yet authors who claim they retain full control over their writing while relying on AI to do it for them are like the social media accounts that gain millions of followers by recycling other people’s jokes and pitching them as their own.

In defending AI “slop,” Marche follows up with this flatulent paragraph:

It is worth remembering that most art has always been slop. When I did my PhD on early modern drama, my supervisor made me read every tragedy written in English until 1640. The process was grueling but instructive. Almost all of them, the fruits of the definitive art form of the greatest period in English literature, were garbage. Go and watch what was on TV in the 80s, if you want a more contemporary example.

Got that? Because Marche didn’t appreciate the zany machinations of the A-Team, writers should be able to make use of sentient calculators. (Marche boasts that he designed a “generative text box” that creates a new prayer every minute. Truly, may God help us.)

It’s true that writing is difficult. There’s a saying that painters love to paint but that writers love having written. As Stephen Fry quipped, “Writing, ghastly at the time but great afterwards, is exactly the opposite of sex.”

So it makes sense that writers who wouldn’t dream of using AI are sore about the ones who do. They’ve put in too much practice time only to have their manuscripts rejected, and lain awake too many nights searching their brains for just the right word, to be rendered obsolete by a clever toaster oven. Writers look at people who use AI to write in the same way that people who endured prolonged dieting so they could fit into their pants look at people on GLP-1 weight-loss drugs: You didn’t earn it.

But there are some mitigating facts to consider. For one, ghostwriting has always existed. That spicy celebrity memoir you loved wasn’t written by a Real Housewife; it was banged out by some poor schlub writing on spec. And nobody really cares.

Further, with AI already unleashed upon the world, there is no turning back. Whether authors admit to it or not, a lot of what you are reading on news sites has been written by Claude, ChatGPT, Grok, Gemini, or any number of other AI services. Spotting AI verbiage will only get harder.

So what happens if AI takes over and nobody even notices? What if Claude had written a perfect movie like Fargo and we found out only after the fact? Would it retroactively dampen our enjoyment?

There’s yet another layer: Large language models (LLMs) learn to speak by using the language and thoughts of other writers who have had real experiences and feelings and put them to paper or pixels. So an LLM will crank out a sentence that may express a real sentiment felt by a flesh-and-blood human in either the recent or distant past. There is an element of genuine personhood buried somewhere in the lexical bouillabaisse that AI cooks up. (The dad jokes Claude purports to have written, for instance, were certainly penned by a human at some point.)

Nevertheless, much will be lost if writing is reduced to simply pressing “Enter.” Even bad human writing has value. (For instance, Marche’s own nearly unreadable column provided the inspiration for this slightly-less-unreadable one.)

It’s a familiar moral problem that some writers, and artists across genres, were terrible people: Can we still enjoy and even love their work? But what if there is no artist behind the art? What are the moral questions then? We cannot question the motivation of AI writing, because computers have none. They have no conscience and no world experience; Claude has never had a drinking problem, or been divorced, or been hopelessly in love. AI does not suffer from self-doubt while writing, and it can’t determine right from wrong based on life experience. It can’t “read the room” when writing for specific audiences. How do we determine whether a sentence has been written by someone with a soul? (For background on what a “soul” is, Claude can probably provide a pretty succinct summary.)

And what could malign forces do with AI? It could be used to create full realities that don’t exist, or express thoughts no human has ever had, causing chaos in a world where AI is our primary way of communicating with one another.

Further, for the more religious among us, AI doesn’t answer to any higher power. It isn’t bound by any guiding principle from above. (This is why churches are asking their pastors and priests to cease using AI to write their sermons.)

The only answer for writers in the age of AI is transparency. If you’re an author who has used AI, then cite it. Otherwise, it should be treated as plagiarism, as most AI platforms are trained using the works of authors who did not consent. And if you're the reader, you deserve to know whether the thoughts and language contained within a column or news piece were generated by an I or by AI.

Let’s try this out: In the interest of transparency, here is how this column was written:

The author wrote it between 10 p.m. and 1 a.m. on the night of Tuesday, April 7, while drinking a Diet Coke and listening to the new album by Canadian band Angine de Poitrine. The author was in a slightly irritable mood, as his Milwaukee Brewers had just lost to the Boston Red Sox by one run. He used Claude to write the first three jokes in the column. He used Google to track down the stories about authors who had been punished for using AI, and to find the stories he had remembered about people using AI to write their wedding vows and church sermons. The author remembered the Stephen Fry quote from having read his memoir, The Fry Chronicles. The author watched The A-Team as a child and thus took exception to Stephen Marche’s opinion of 1980s television. Further, the author recently listened to a podcast about Fargo, so the movie was at the front of his mind.

The author ran the piece through an AI detection tool, and none of the text showed signs of AI.