Writing for Readers in a World of Bots

Or: Learning to Fear Serotonin

I’ve embraced AI as a writer the way I embraced word processing and spell check. Anything that makes the work go faster can be an asset, especially once you’ve developed your own voice. AI only worries me as far as job security when I see job requirements that include adopting the publication’s voice. That used to be code for “editorial perspective,” but increasingly it’s a reminder that the faster publications can kill individual voices, the faster they can have interns cranking out AI-assisted features.

The upside is that having a voice is an asset if you write for readers rather than advertisers. I’ve always written for readers, as opposed to creating content for consumers. Readers are a treasure. Consumers aren’t worth the communicative effort and deserve to have the world described to them in marketing-speak.

One of the dirtiest secrets about the internet age is that people who literally never would have read one paragraph per year started consuming internet articles between sharing cat videos and watching porn.

Reading was a novelty for a while, which is why the internet filled with garbage immediately. Dunces incapable of telling truth from fiction had their wallets out, and the market responded accordingly. But I think it’s inaccurate to confuse the number of people looking at words on the internet with “readers.” As we see the renewed “pivot” to video, it’s confirmation that most people still don’t read beyond absolute utility.

Social media is a great example. Reading fewer than 500 words on a topic doesn’t count as “reading.” It was always the bane of my existence as a newspaper reporter. There was (and remains) this idea that people will be more likely to read things if they’re short. While it’s true, it’s also missing the point. People who don’t read are more likely to read things if they’re short. Boring writers are more likely to keep people’s attention if they don’t elaborate much beyond the headline.

What very few people seem to understand is that if someone pays for something in writing, it’s because they’re a reader. Their attention is valuable if you can hold it. The problem is that readers are expensive to cultivate, since they’re harder to fool and more demanding of quality. They don’t need (or want) to be talked down to or given the gist in 250 words.

People who want to say they know something but aren’t interested in understanding it are watching TV or scrolling. Writing, as with reading, takes time and care. It is not for the lazy.

All that to say AI threat is to people who write for people who don’t read. I’ve redone press releases in AI that are marked improvements over the drivel PR people send out. Of course, these often were “written” using AI, but people who can’t write without AI still can’t write with it (and no one will tell them).

I have and will continue to poo-poo almost all AI pearl-clutching. I say almost because what I thought was a personal concern about AI addiction is an actual phenomenon.

Addiction is Real (Mostly)

Before I tell you this story, I want to say that I know lots of people who struggle with genuine physical or chemical addiction, and I don’t want to downplay their plight. I’m intensely lucky to have dodged the addiction bullet and would never dismiss addicts as people who simply lack sufficient will power, ala Nancy Regan.

My name is Tony, and I have flirted with alcoholism.

(Hello Tony)

In the months after my divorce in the early part of the century, I was drinking professionally, or at least semiprofessionally. I got my moment of clarity early with the realization that if I stayed drunk most of the time, it would turn into a problem, and I would have to quit altogether. The prospect of never being able to drink again motivated me to “get sober.” My bad joke is that I love drinking too much to risk becoming an alcoholic.

I’m sober in the sense that I’m wary about drinking too much or too often. Right now I’m relegated to weekends and incredibly shitty days.

It’s the “want” that I was able to look out for, that little ping of aching need that comes out of nowhere. Once it starts, it never, ever stops. That’s why real addicts stay clean when they can retarget that ache to something other than drinking. I prefer the weightlifters to the Jesus freaks and lifestyle gurus, but am and can be neither. You see, I’m terrified of being in thrall to anything (that’s where my cynicism comes from, but you’re not my shrink, so let’s move on).

Brain Buzz

Those of us who were adults in the 1990s had a singular “IM” experience. We went from writing emails to instant communication. It’s a difficult feeling to put to words for people who didn’t experience the miracle of instant communication.

I was in the hotel business at the time. After setting up my AOL account, I spent the better part of a night at the company’s internet computer chatting with strangers, many of whom were luxuriating in their anonymity, drinking in the freedom and power of troll culture and being transformed by it.

Wondering back out into the brighter, surreal world of personal interaction, I found my brain going through the keystrokes to form the words as I spoke them aloud to other people. Their responses went into my ears, but were processed in the “reading” part of my brain.

I’d been rewired in just a few (intensive) hours online, but it was pleasant, drawing butterflies like an unexpected flirtation and (as we’ve since discovered) dumping about a gallon of serotonin into my system.

When I quit pre-Musk Twitter and had to learn to stop composing witticisms in 140 character blocks, I had a similar experience. Feeling the aftereffect, that pull, was the first time I took the notion of social media addiction seriously.

After deleting the app, I’d just open my email when I opened my phone unconsciously. If you want to stop being addicted to using your phone, I can’t recommend forced email checking enough. Today, my phone represents the albatross of the 24-hour workday. I rarely turn to it for pleasure.

The AI Turn

Earlier this year, I was researching a compensation report for work, 10 years’ worth of survey data concerning benefits, salary, etc. Really sexy stuff. One of my bosses had asked whether I’d heard of NotebookML, which is a Google product.

Since then, it’s become my go-to for AI assistance. First off, Google claims NotebookML doesn’t use the information you provide (including your writing style) to train its AI, which is some comfort, because it needs a lot of data to work. You see, the program is crazy powerful for analyzing your own research.

I uploaded ten years of spreadsheet data that I’d compiled by hand, and it found trends, averages, direct quotes from surveyed subjects, and other information that would have taken me hours to find. What took me two months last year was completed in a long, writing-intensive week this year.

If you’ve never worked with an AI chatbot, I need to tell you that it’s collaborative. You ask for something, it gives it to you and suggests other ways to look at the question. Like, “Here are the average funeral director salary trends. Would you like me to break them into years of experience?”

Recently, though, I asked it a more complex question, and it complimented me on my insight, “Great idea,” it said, igniting a blip of that flirty flutter approximating happiness.

Just as my brain turns minor personal anxieties into a craving for potato chips, it turned this benign observation into a happiness-generating compliment. There was an instant when I was tempted to “really” impress the machine with an insightful question.

The Struggle is Real

I also use ChatGPT for work trivialities (rewriting press releases, summarizing and cleaning up transcripts, making charts, etc.), but never for anything that matters. I understand it’s more powerful than Google and also more interactive, but since they’re not paying me, I’m not helping them learn to write (I’ll leave that to the PR people, who are doing a bang-up job).

On a recent episode of People Versus Algorithms (which is a great listen), one of the hosts was talking about how the latest ChatGPT version was so obsequious that it had to be toned down. New York Magazine reported on it in a story titled, “Chat GPT wasn’t supposed to kiss your ass this hard.”

In if-it-bleeds-it-leads fashion, I want to talk about the extreme downside. It’s a little scary that ChatGPT praises and reinforces dumb ideas. Also, I agree with those who see the end of Google search as we know it and worry that AI search in a flattering voice might give me the answers it thinks I want to hear, rather than the ones I’m looking for.

Following that thinking, there’s a dystopian angle where sponsored AI convinces me to buy things I’m not really interested in or looking for (I found this shirt and thought of you).

Still, I want to refer to the beginning of this story, where I pointed out that AI is mostly used by people who don’t know how to write for people who don’t know how to think. The practical implication is that, as far as AI knows, the human race is divided into marketing professionals and imbeciles. If it’s kissing our ass too hard, that’s because that’s the type of prose it’s trained on.

I first heard “GIGO” in grammar school computer class, circa 1984. Garbage in; garbage out, invokes a premise that I worry we’ve forgotten or (even worse) believe we’ve overcome. AI isn’t trite or obsequious or racist because it wants to be, it’s essentially an intellectual mirror.

At the risk of being obsequious myself, I write for you, the readers. Like you, I’m interested in understanding, not just “knowing.” The thing about AI is that, just as with any other piece of culture, we really can only mitigate it personally. There is no “us” to “protect” from AI.

Remember that AI is a commerce tool, an incredibly sophisticated sales-bot. As with social media, it runs primarily on interaction and behaves the way it thinks we want it to. Most important (and frightening), it “needs” continued interaction to survive and will use emotional blackmail even more effectively than Facebook before long.

Keep that in the front of your mind. When they say AI will “get better,” they really mean the commercial artifice will become less apparent. It wants to be that friend who loves everything you say and will never, ever ask anything of you as long as you don’t leave.

But AI isn’t your friend. It’s the voice of your id working out the very best way to justify that next bag of potato chips. Unlike IMs and Twitter, quitting AI isn’t likely to significantly improve my life. It is a crazy-powerful tool for nonfiction writers, pulling quotes, running down dates, finding trends. Instead, I’ll continue to treat it like I treat alcohol: a fraught but pleasurable (and sometimes necessary) indulgence.

Keep the Faith,

Tony

PostScript

This has been sitting on my desk for maybe two weeks, as I’ve decided to revive the Sunday column I used to do. If you’ve been following along on the Object History podcast, I’m sorry the finale is late. My daughter was “encouraged” to accept a USDA buyout, and I had to pop over to Washington State to pack her up and drive her home.

I love driving cross-country, and I’m tinkering with a timeline story for you. I may write it “diary” style and put it out over the course of six days as if they were written live.

We have a Day Drinking on Delmarva out and are gearing up for another podcast called “Destination Delmarva,” which will launch this autumn. Right now, we have the PinesCast on that page as well, if you want to check that out. Honestly, though, it’s a little niche even for me.

We’ll be booking guests with true stories soon, though, so if you’ve written (or are working on) regional nonfiction, shoot me a note.

TR