Of all the slings and arrows coming for journalists these days, the one we honestly didn’t see coming — until all of a sudden it’s smacked us in the face — was ChatGPT. A.I. is progressing so much more swiftly than even its makers imagined. Only two years have passed since The New York Times astonished the newspaper-reading world with a feature detailing how ChatGPT-4 could draw a clumsy unicorn when prompted and, as of this moment, elementary school kids are creating deep-fake videos and feature-length animated films simply by asking generative A.I. bots to do their bidding.
And, yes, freelance writers are turning in stories to weekly newspapers composed in the smooth, dulcet tones and sweetly swinging rhythms of ChatGPT-generated text.
Here are a few clues that that “news” item or magazine feature you’re reading online was written by A.I.
First, it will be smooth as cream and include uplifting, inspiring sentences that ring with faux sincerity. It will use lots of words like “empowering” and employ the singsong cadences of the pulpit or a Hallmark card, working in phrases such as “not about competition but about caring” or “not about winning but about beginning.” ChatGPT is ever so earnest.
Second, it tends to use distinct sentence structures that, to anyone who has spent an entire career reading human writing, ring out clear as a bell. It piles on so many em dashes that high school kids have become aware of its em-dash addition and, even as you read this, are feverishly working to remove em dashes from their A.I.-generated school essays, like so many cat burglars trying to erase their footprints in the snow. (Woe to those human writers who have always loved a good em dash!)
Third, it will compose texts in “triplets.” All things come in threes in the ChatGPT brain, and so, employing the Oxford comma, it will not just encourage the reader to “dream big, work hard, stay humble,” but every example provided as evidence for an argument will come in threes. If the make-believe ChatGPT reporter were to submit an article on the migration of geese, it would say things like “Their path is ancient, their pattern precise, their purpose perennial.”
Finally, ChatGPT texts come in polished to perfection. There are no quirks or hiccups, no flavorful but perhaps not quite relevant human asides, no mess, no run-on sentences. It uses sentiment and therapy-speak to tug at the heartstrings but it lacks variety. It contains no phrases or ideas that, to borrow a phrase from the poet Gerard Manley Hopkins, are “counter, original, spare, and strange.”
The Star has a few wizened, seasoned text editors around our office. Imagine us wearing hats with “Press” tucked into the band: Collectively those who edit submissions have on our résumés perhaps a century of reading the work of human beings, and, well, guess what? We can tell.
The question is: Will younger generations become familiar enough with human writing to know the difference, or care, or, indeed, to prefer the written word as generated by an individual brain rather than an algorithm that is smarter than us?
At the rate things are going, we expect a Pro Humanitate movement, calling for a return to human writing (with all its flavor and failings and human observations, whether peevish or awed) by 2026.