On Doing

Good writing, AI, and superficiality.

In the interest of professional development, I had ChatGPT read a few of my recent essays. Beyond stroking my ego with such praise as "[having] a strong voice, engaging storytelling, and meaningful themes,” Mx. GPT also had some constructive criticism.

Apparently, I need to work on my “pacing & sentence structure” and should “experiment with short, punchy sentences for contrast.” My narrative stakes are low: “some of [my] more contemplative pieces could use a bit more… energy to keep readers hooked.” Oh, and my essays “feel more organic/free-flowing than structured,” and could stand for “more deliberate openings and closings” to “amplify impact.”¹

I suppose those things are true.

I like writing — not just the physical act of typing out my thoughts, but also the technical challenge of putting them down coherently on the (proverbial) page. I’ve spent a lot of time thinking about things like syntax and grammar and the finer points of style — and in doing so, I can, with relative ease, produce what I like to think is good writing.² But enjoyment does not make writing easier.³ To write well, one needs to draw from experience — both personal and collective — and effectively transmit that through to the reader. Like any craft, good writing takes work — some skill, some patience, an understanding of technique and the willingness to make an effort.

For most of history, too, good writing required more than skill — it required access. It required finding a way to be read. There’s the Zen koan: “if a tree falls in the forest, and no one is there to hear it — does it make a sound?” Good writing is much the same way — if it exist only in the personal diary of the one who wrote it, is it really any good? Thanks to the internet, publishing your own writing is now as easy jotting a few words into a word processor and hitting “send.” Personal diaries have become public. I can transmit this essay to all six-hundred some of you with less effort than it takes me to walk up the stairs into my office.

Still, though, rapidly writing an internal monologue and distributing it to the world does not good writing make — it takes mental effort and proficiency to craft well-written prose.

The moat is shrinking, though. Using language models like ChatGPT, anyone with a spark of creativity can workshop their ideas into an The New Yorker quality essay with a few prompts.⁴ Or why even bother with that — let the model know a few authors you admire, give it a theme, and then have it spit out an original essay for you. It won’t even mind if you take full credit for writing it.

I did that, out of curiosity — asking ChatGPT to write an essay in my own style. It did so almost instantly, cranking out an essay reflecting on tools “only being as good as the work that they do” and using a knife (my knife!) as the underlying metaphor.⁵

I’ll be honest — the writing was good. The sentences were crisp, yes, but it worked in enough of my own meanderings to produce a facsimile of something I would write. The topic was on point (even if something I’ve written about in the past)⁶ and it managed to make appropriate, albeit creepy, references to Kiddo and the Warthog. I may steal from its ending to make my own closings “more deliberate.”

The essay lacked something, though. I’m loathe to call it the “human spark” because that would be cliche and also not entirely true. The essay did feel quite human. It was technically flawless, used varied sentence structure, began with a strong opening paragraph and closed itself out cleanly. Selfishly, I’m proud that the model was able to generate it based on my own writing; truthfully, sadly — even frighteningly — I’m not entirely sure I can pinpoint what was missing. But, however proficient — it was still missing something.

In its essay, the model wrote that “[a] knife that stays too sharp, too perfect, too pristine, is a knife that isn’t doing what it was made to do.” That “[the] things we treasure most should bear our mark.” The metaphorical knife in that essay “is not perfect. The edge is a little dull, the steel blotched where blood met blade and wasn’t cleaned quite soon enough. But it is a good knife. It has done things.”

There lies the problem with “good” writing produced artificially. In a twist of irony, that writing is perfect, it isn’t dull, it hasn’t done things. It doesn’t bear our marks. And I suppose — at least for now — we should take solace in the fact that that doing is something that only we, us humans, can do for ourselves.

We’ll see you back here next week with some writing rooted in reality and some food to support it.

1. “Maybe start some pieces in media res (right in the action) to hook the reader faster.”

2. And I’m glad that ChatGPT agrees.

3.

Lawson Chapman

wrote on this topic a few weeks ago.

4. When I asked ChatGPT to review this essay, it suggested that “The Atlantic’s Ideas section would eat this up.”

So — editors at the Atlantic, if you’re reading this…

5. I wrote this essay before I wrote last week’s dispatch about… knives. And

Sam Alaimo

mentioned in the comments of that essay that he’s working on a piece about… knives. A classic case of life imitating art imitating imitation life and imitation art!

6. Though about cast-iron pans and I did not give that essay as an example for it to read.