🌲 AI cannot replace genuine human connection

That doesn't mean it's not incredibly useful, or economically disruptive, of course.

New year, new me! I renamed this newsletter and swapped some images out. New logo — did you know 'Konik’ means ‘pony’? Anyway, I updated the About page. There’s now a statement of how I use LLMs (more on that soon). I also added a link to my YouTube channel, which I am dusting off for Thursday morning livestreams. Expect the format there to vary over the next month or so as I figure out what works — 3mph on the treadmill while reading and taking notes was a bit much…

---

My timeline is flooded with people talking about the new AI models — and Deep Research seems genuinely impressive. For myself, tho, I’m more focused on how DALL-E is finally good enough that I cancelled my Midjourney subscription. I’m not sure if the model improved or I got better at prompt engineering, but the last few “featured images” for my articles have managed to avoid the annoying “GPT glow.” I now have a “GPT” called “Artist” that does a decent job of creating images that do a much better job than my old stock photos at representing what my article is about.

Weirdly, it doesn’t let me select which model I want to use with the prompt. As impressive as OpenAI’s many offerings are, I’ve noticed they often feel a little disjointed. This one can use the internet, that one lets you upload files, but if you want it to remember your prompt for a later task? Too bad!

Keeping up with AI models definitely feels like dancing on the jagged edge of technological change. Sometimes, in self-defense, I retreat to my paper books and my analog notebooks, my walks and my garden and my home-baked bread. That balance is only to get more important as the years go on, I think.

Other times, I dive in so I don’t fall behind. I’ve always liked finding ways to solve problems with maximum efficiency.

I now have a series of saved prompts in my Obsidian vault tucked into my Templates folder. One has a really detailed explanation of what my technical background is, to help with debugging weird phone problems. I’ve noticed, for example, that I get better hints if I tell it I work for the company who makes whatever software I’m having problems with. If I omit that, it tells me to contact customer support. If it thinks I am customer support, it usually has a troubleshooting step to offer I hadn’t thought of.

Did you know that Boox Go Color 7 apps don’t show up in custom launchers if the apps are “frozen,” even its “frozen” status in no way kept the apps from launching or functioning fine. 20 minutes of troubleshooting later, my nighttime reading sessions involve a lot less cognitive load getting to my book. Thanks, OpenAI!

Anyway, here’s “Artist” —

This GPT is designed to create book cover art and featured images for EleanorKonik.com. It will generate stylized vector line art digital paintings with a 5:8 aspect ratio for cover art and 3:2 for newsletter art. The color scheme should primarily focus on muted, washed-out maroon and teal. Women in the illustrations, unless otherwise specified, should generally resemble 35-year-old blonde moms similar to the women in the provided images. The goal is to produce artwork with a dynamic, bold, and vibrant feel that suits the site's visual aesthetic. Use lots of negative space in the background, be very minimalist, very flat and matte.

Although I personally have never paid for art to use as featured images (Before AI, I always used images in the public domain), the widespread availability of this kind of thing is obviously going to have impacts on the economics of visual art.

It might even one day get good enough to create consistent art for comic books, although it is definitely not there yet. Ursula Vernon’s 2021 observation that “sooner or later someone is going to notice that nothing takes place in the same scene from a different angle” pretty much holds true — in that experiment, she personally drew the figures, but even today it’s very difficult to get an AI to spit out art that looks like it’s from the same character, or to get all the details of a specific character exactly right.

The day might come, but even if it does, I am going to continue to buy Ursula Vernon’s comics. Why? Because I have a parasocial relationship with Ursula Vernon. I have been following her work since the livejournal days, back when Digger was just a quirky webcomic, before the lolwut pear went viral. I like the way she thinks, her life interests me, she is a real person, with a mom she loves, who is funny and decent and has silly dogs and a big garden I’d love to emulate.

It helps that I respect her for staying on top of AI trends and sharing their progress and how she can (and can’t) integrate AI into her workflows.

Some people decorate their houses with art that matches a color scheme, or the vibe of their room. I’m not one of them; most of the art on my walls has some kind of meaning. Meaning is maybe the one thing AI can’t find for us; I don’t care how much market share AI girlfriends companies are grabbing in the current boom. Absolutely nothing made of silicon can come close to even a pale shadow of the relationship I have with my husband, my children, friends I would move mountains for even when I want to wring their necks for being so obviously wrong about whatever thing we’re having a good-natured fight about this time.

AI cannot create a family heirloom. An old aerial photograph of my father-in-law’s (long since sold) family farm. An oil panting of a winter landscape painted by my husband’s great (great?) grandmother. Inkblots of the Tardis and the Serenity I picked up at a local science fiction convention. Desert paintings I picked up at a company offsite in Santa Fe for my husband, because he spent his formative childhood years in Albuquerque and loves the region. The seascape I got during senior week “downee ocean.” New York street art my husband got in his twenties. A Sailor Jerry pinup girl print a friend gave me as part of a joint gift with my then-boyfriend — the ex got the rum. The baseball Jim Palmer signed twice, damn near twenty years apart.

It is literally impossible for AI to re-create the weight of that history. This does not make AI useless but it does help me cope with the occasional feeling of “why bother?” when it comes to writing, or reading.

Right now, I am reading Black Lamb and Grey Falcon by Rebecca West instead of a Wikipedia article about the history of Yugoslavia because it gives me a sense of genuine personal connection with another (long dead) woman. I read the Honor Harrington series instead of an Oxford deep dive into the history of the French Revolution and the Napoleonic Wars because it’s an intimate reflection of David Weber’s political opinions and gives me a reference point for understanding the French Revolution and the Napoleonic Wars.¹ But also because it’s fun to read. It’s human in a way that scholars often don’t manage, due to the constraints of academia (which Erik Hoel and Étienne Fortier-Dubois have touched on multiple times).

Certainly AI can mimic anyone’s style — but it mostly cannot be relevant the story the way that the history Patrick McKenzie and Matt Levine and Jane Psmith bring to the table matters. I have had actual in-person conversations with authors like Tom Doyle² and Ada Palmer³ in a shared social context; those conversations add texture to the experience of reading their books that large language models and robots cannot, by definition, replicate.

There is a time and a place for getting answers from the id of humankind. My evening reading time (last night’s book was The Silk Roads by Peter Frankopan, yes I do enjoy reading things written by Byzantine scholars) ain’t it.

The technological landscape is changing rapidly — but that’s been true for a long time. Longer than you might think. Anatomically modern humans have been around for 200,000 years, spoken language probably 50,000 years ago, and the state is only about 5,000 years old. The industrial revolution feels like a long time ago, but on the scale of human history? We’ve been living in a time of extreme change our whole lives.

We’re still people. Despite the printing press, despite the telegraph, despite cell phones, we’re not that different from our grandparents, or even our earliest forebears. Maybe there’s a fertility crisis, maybe there’s a population crisis, maybe there’s unprecedented war and maybe there’s unprecedented peace, but for myself?

I hope to be reading — and writing — essays for another fifty years, out of my own mind and with my own hands. I hope my friends who paint keep painting, too.

1. (Shut up LanguageTool, I don’t care if “shorter sentences make the text easier to read.” Not one single essay I’ve enjoyed reading in the last fifteen years has been easy to read.)

2. I like to describe Tom’s American Craft series as “A Baen urban fantasy book… as published by Tor” but the actual blurb begins: “US Army Captain Dale Morton is a magician soldier—a “craftsman.” After a black-ops mission gone wrong, Dale is cursed by a Persian sorcerer and haunted by his good and evil ancestors.” It’s a fun romp through the idea that early American history might have left us a magical legacy; rather like A Discovery of Witches but for American men instead of British women. Here’s an affiliate link for the first book, American Craftsmen.

3. Ada Palmer is the author of the incredibly complex and thought-provoking Terra Ignota series, as well as being a professor at the University of Chicago. Terra Ignota is a near-future sci-fi story that takes thoughtful aim at the cultural gestalt of its time. From the blurb: What seem to us normal gender distinctions are now distinctly taboo in most social situations. And most of the world's population is affiliated with globe-girdling clans of the like-minded, whose endless economic and cultural competition is carefully managed by central planners of inestimable subtlety. To us it seems like a mad combination of heaven and hell. To them, it seems like normal life.