An occasional letter writer to The Star sent us an addendum to a recent submission last week about artificial intelligence. After writing two short lines complimenting a front-page photograph by Durell Godfrey from last week’s paper, Ora McCreary tried an experiment; she pasted the draft into ChatGPT asking for any comments or suggestions. Speaking in the first person, the chat-bot declared it would be happy to help. “Please go ahead and provide me with the draft, and I’ll provide you with any comments or suggestions I have,” it said. So Ms. McCreary did.
ChatGPT took things a lot further, however. Instead of comments or suggestions in the style of mere spell-checks, it rewrote the letter entirely, from start to end, and it piled the hyperbole on thick. Instead of loving the photograph, as Ms. McCreary originally wrote, ChatGPT “wanted to take a moment to express my admiration” for the picture of “the high school graduates, brimming with hope as they gaze towards their future. . . .” As for the photographer, “Their skill in capturing the essence of the moment is evident . . . their ability to evoke such emotions through their lens.” And on it went, “I commend your publication . . . dedication of your team . . . a valuable source of inspiration and information for our community.” Blah, blah, blah.
In an end note, ChatGPT explained that the new letter “maintains the overall positive tone . . . while expanding on the reasons for the author’s admiration.” Oh, really? Ms. McCreary went with her original, rather than the cloying, sugary rewrite. It was all very amusing, except for its implications for human communication — and the keepers of letters-to- the-editor sections.
The thought crossed the mind after receiving Ms. McCreary’s report that the dreary sameness in our recent storm of letters from the Maidstone Gun Club might have had something to do with A.I. — certainly, they were universally grammatically cleaner than the almost-every-line corrections that are often required in letters. And, for example, election season letters are already difficult enough to slog through. A.I. could make things much, much worse.
We get to know our regular letter writers by their quirks: one who uses commas in place of spaces, the Random Capitalization guy, the queen of run-on sentences. Already, we suspect that apps like Grammerly are taking the variety out of preparing each week’s batch for print. We will miss them if they switch to asking ChatGPT to argue their point instead.
Then, too, we wonder how newspapers and high school teachers and human resources departments will cope with the coming wave of nonhuman writing. The only solution would seem to be requiring hand-written submissions only — but then again, they could be copied from what some computer somewhere decides expands on what it believes the author had intended to say in the first place. The solution perhaps is to let the machines read the machines — and leave us living, breathing entities out of it.