Before Going Off-Grid for a Few Weeks I Offer a Few Words About AI.
[always free]
1.
I want to start with a line from a short story by Steve Almond1 I quoted in a 2024 post [Surprise Me]:
If absence makes the heart grow fonder, ultimatum make the heart grow feral.
That post was about your basic job as a writer: Keep readers reading. And what the kind of sentences you write has to do with making that occur. You want the reader’s mind to be awake/alert/expectant, I said, which won’t be the case if you write dull, predict-able, voiceless sentences. You need to surprise people often enough that they know you’re capable of knocking them off center (as Almond does with his final word).
Today’s post is also about surprise . . . but I want to connect it to AI.
2.
Daily life relies on outcomes matching expectations. You feed your quarters into the machine, press B6, expecting the bag of Peanut M & Ms to drop, and it does, nearly every time. Yay.
Storytelling works off a different model: storytelling is rooted in anomalies. As a story begins, we listen for a certain paradigm, some version of: Usually . . . but today. It’s what tells us we’re hearing a story.
Further, good writers know that reading is linear, that stories unfurl in time, that you don’t spill all your beans at once, but feed them out artfully, letting curiosity/tension accumulate. In shows like Law & Order, we expect a plot twist—a final dollop of information that upends everything.2 But the same idea—holding back, then releasing—applies to the story’s smaller structures as well—subsection, paragraph, sentence.
That was the idea behind this question:
Let’s say you’re a writer. And let’s say some troublemaker comes along and puts a thumb over the last part of the sentence you’ve just written. How often could someone else figure out the covered part?
My point being that there’s a kind of writing that tries not to confuse, not to offend, not to stick out, not to actually say anything, but instead to leave a good impression—Chamber of Commerce prose, I call it.3 In such writing, you can guess the hidden part because it’s the same stuff the sentence started with. There’s no treat awaiting you, no silver coin for your metal detector to beep at. And so, if you’re the reader, you get drowsy, your gaze jumps ahead to the next paragraph, you wonder, How many more pages of this do I have to get through??
How easy it is to read over the top of this writing—the emotional blandness, the monotony, the poverty of the word bank . . . there’s nothing to slow you down, nothing to excite the eye or mind, nothing to jar or snag you. I remember, years ago, stumbling onto this little paradox: Here I am line-editing, snipping out extra words, clarifying, correcting the punctuation . . . If this were woodworking, I’d call this step sanding down the rough spots. Yet, here, I don’t want a glossy surface, I don’t want my reader to zip across it as if in stocking feet. I need to roughen up the surface, I need to make the rhythm twitchier, I need more sharp objects, more collision, more attitude.4
3.
Now, why am I rehashing this stuff? Because, a few weeks ago, I flew back East for my [redacted] high school reunion, and heard the new head of school’s thoughts about AI’s impact on the curriculum.5 As she spoke, I had what felt like an epiphany about how AI goes about its work, vis-à-vis how writers do.
As an experiment, I asked AI to finish Almond’s line—and since Almond’s tone was edgy I asked for “sharp” replies.
AI’s responses:
If absence makes the heart grow fonder, ultimatum makes the heart pack its bags.
If absence makes the heart grow fonder, ultimatum makes the heart lawyer up.
If absence makes the heart grow fonder, ultimatum makes the heart swipe left.
If absence makes the heart grow fonder, ultimatum makes the heart change the locks.
Then AI said that, depending on the specific situation, common alternative completions were:
If you’re resentful over being forced into a corner, you could say:
. . . ultimatum makes the heart grow bitter.
If the moment’s characterized by emotional withdrawal and self-preservation,
. . . ultimatum makes the heart grow cautious.
If the moment expresses the psychological detachment before a final breakup, you could say:
. . . ultimatum makes the heart grow distant.
If you’re expressing exhaustion from ongoing relationship tension, you could say:
. . . ultimatum makes the heart grow weary.
And so on.
Later, AI asked if wanted to “expand the quote into a full stanza or poem.”
Um, thanks but no thanks.
4.
Here’s what I realized (but first I have to admit I’m so little schooled in the AI universe that I don’t know whether this is truly an insight or something everyone already knows):
Ask AI for help, and, lickety-split, it scans . . . not everything ever written, but an impressively vast trove of what’s accessible, flagging the hits, which it aggregates, organizes, and gives you in a format that mimics a helpful friend’s advice.
In certain sports (for instance, figure skating, if I remember this right), the protocol is to throw out the judges’ highest and lowest scores, then to average the rest, and that’s your final score. In other words, it erases the outliers and calculates the statistical middle. Isn’t this what AI does? Its highest yardstick is: What gets the most hits?
Whereas, my yardstick is, to put it bluntly, What meaningful thing can you say that no one has ever said before?
My guess is that, in the history of the universe, “If absence makes the heart grow fonder, ultimatum makes the heart grow feral” has been written only once (discounting all the times I’ve quoted it).
Which is why it would never occur to AI to suggest it.
5.
Today’s post was also instigated by fellow Substacker, Victoria Livingstone’s comments in “Texts That Slither: AI in Granta,” at her page Human Generated:
Yesterday I read “The Serpent in the Grove,” one of five winning stories selected from 7,806 entries for the Commonwealth Short Story Prize. The winning stories were published in Granta, one of the most prestigious literary magazines in the world.
Her assessment: I hated it.
She proceeds (passing on an idea from a former professor):
. . . certain texts offer the illusion of cohesion. If you look at individual paragraphs, they are well written and seem to be making some kind of point. Transitions, too, seem smooth. One paragraph slides into the next. However, when you step back from the article, you realize that there is nothing to hold onto. The entire argument seems to unravel, or perhaps there was never anything solid beneath those phrases that so smoothly snake down the page. That is how I felt reading “The Serpent in the Grove.”
And then says this:
Some paragraphs overdo the use of simile. Take, for instance, these lines:
Sita moved quiet as if sound were taxed. Nineteen and brown like dust after rain, she turned roti dough with a rhythm that came not from joy but from endurance. Orphan was too kind a word. Orphans are sometimes cradled. Sita had been passed like a parcel from kin who were hungry for everything except another mouth; she learned to make herself small.
Three separate similes, all working with different imagery: “as if sound were taxed,” “like dust after rain,” and “like a parcel from kin.” Also, “dust after rain”? Isn’t that just called mud?
6.
I found an echo of this in an Iranian novella I read this week, Shahrnush Parsipur’s, Women Without Men (1989).6 In her author’s note she says:
My cousin . . . decided to become a poet at the age of thirty-seven. Becoming a poet has become a common practice in Iran. People without knowing anything about the rules of poetry, put words together abruptly and, using weird thoughts, believe they are creating poetry. For example, “Light’s affection is running in electric wires,” or “The scream coming to the surface of existence was violet in color,” or “Earth’s Red told the Blue of Presence: I don’t like destiny.” And so on. Some of these poems are interesting, but become ridiculous when, in order to cover their own illiteracy, some poets claim that the grammatical conventions in poetry are nonsense and have to be discarded altogether.
Don’t these faux similes read like AI’s?
7.
When I was teaching, I sometimes explained a lesson I took away from my first (and only) backgammon tournament. The quickie version: It was a small event and after two rounds I found myself playing for the championship. I’d read the books (such as Backgammon for Blood) but was essentially a noob. My opponent was not a noob. But he kept making moves that flew in the face of what the books said. He won handily, of course. Driving home, I began to parse his strategy. Every turn in backgammon starts with a roll of the dice (the luck part), followed by advancing one or two of your pieces (the skill part). In short games, luck can dominate. Longer games favor sophisticated decision-making skills. Therefore, my opponent made sure the games were messy, complicated, because he knew I’d be out of my depth.
Now, how this pertains to writing: You don’t have to start with a wholly original idea. In fact, waiting around for one is a time-waster. “The way to be original is to be yourself” . . . sage advice attributed to various artists/writers. But it’s incomplete.7 I’d add (my take-away from the backgammon tournament): You need to make the process decision-rich. Think of the instruction not as a vague encouragement, but as a description of how good writers get from general to specific, anonymous to personal, stale to fresh. Think of “yourself” as a filter you pass your material through—do it enough times and you end up with an artifact no one else could’ve written. More decisions = less “anybody could’ve written this.”
Which brings us back to AI.
I was working on a joint project with a friend and he used AI to aggregate responses to a questionnaire. AI was good at that. It addressed my friend by name, and, like a helpful co-worker, ticked off a few related tasks it could do if asked (though it stopped short of offering to write him a poem).
Several times, I’ve had to ask Substack Support a how-to question; we’ve chatted back and forth, zeroing in on what I wanted. It wasn’t hard to keep up the illusion that I was conversing with a human call center worker. But remember what Livingstone said above . . . when you step back from the article, you realize that there is nothing to hold onto. There’s nothing to hold onto because there’s a void at the center. AI can pop out zany similes, but it can’t create the kind of unique observation that springs from the innumerable scraps of lived life cached in a writer’s brain. The dictum: The way to be original is to be yourself has no meaning for AI. It has no self.
8.
In June, The Republic of Letters re-posted a piece [Raise High the Banner for Literary Standards! A Cri de Coeur] written by Tyson Duffy at The Literarian Gazette. I recommend both sites.
Birth Year Project:
Link to most recent post: https://longd.substack.com/p/birth-year-project-1973. Still open for business.
OK, good summer to you, one and all. See you the first week in September, by which time I’ll have read East of Eden and My Brilliant Friend (he said, going on the record).
Steve Almond: From “The Body in Extremis,” in My Life in Heavy Metal (2002).
Law & Order: I remember an episode where the trial ends satisfactorily and the team heads for the tavern; we start to relax . . . but, uh oh, a casual look at our watch says there’s ten minutes left. Bum-bump. And maybe you’re old enough to remember the M*A*S*H where practical joker Hawkeye waits all day for the payback he deserves, suspecting every transaction to be the one that gets him, but instead he gets the anti-payback, the day-long wait where he has to keep proving to everyone he’s onto them, they’re not gonna fool him. But they already have.
Chamber of Commerce prose: Or Eddie Haskell prose.
I need more: . . . words like twitchier.
New Head of School: This took place in the same auditorium where I once played Mr. Bumble in Oliver, sporting a bald wig that looked more like a sepia-toned beanie.
Women Without Men: I’ve been prepping a post on Iranian novels that will appear early in the fall.
Sage advice: Sometimes we give out writing tips that make sense to us without realizing we haven’t made the advice sufficiently granular. Cut extra words, for instance. But which words are the extra ones? Line editing is the easiest part of writing well, but you need to learn how tightening prose works. Several times, magazine editors asked me to shrink stories I considered finished, fully edited.
The first trick was to remove as many individual words as it took to shorten each paragraph by a line. In part of an hour, I saw, you could make a 25-page story a 24-page story, without sacrifice. Then I looked at all the qualifiers—adjectives, adverbs, hedging expressions like kind of/sort of, etc. I didn’t need many of them. He had an old man’s long droopy ears. And I could, often, say something more directly, less grammatically roundabout. I’d scour the manuscript exclaiming, Oh, that can go! And that! Only toward the end of the process did I lose any vital tissue—a scene, a couple of pages of back story. When it came time to reprint the story in a collection, I used neither the original 32-page version, nor the stripped down 24-page version; I put back the chunks, but kept the bulk of the edits. The 26-page version read better, sounded more confident, more decisive.





I will miss you while you are gone. I wanted to send you a copy of my novel, coming out next week. I have an address beginning 49 that I hope is still correct.
I love the Chamber of Commerce comparison! A lot of AI writing is like that; peppy, encouraging, always looking for how best to be supportive.