Five years ago, I spent perhaps 30 hours reading Seneca’s letters and rewriting them with curse words, a task an AI could do in seconds. Shall we just die stoically, then?
By Jennifer Dziura
16 min read
A post to X on Nov 20, 2024 smugly reported that, in an online study, “AI art haters were unable to distinguish AI art from non-AI art.”1
The post referred to the “AI Art Turing Test,” a study devised to determine whether people could tell AI from human-created art (spoiler: not reliably), and which they liked better.
The report about the “AI Turing Test”2 shows a long series of paintings (or, in some cases, “paintings”), challenging us to determine which are AI, to choose our favorites, and then to view the results.
You can take this test yourself, and just like the people in the original survey, you, of course, will be wrong sometimes. A lovely painting you are sure was created by some long-dead Frenchman was in reality created by a teenager in a hoodie sitting in a padded gaming chair using his killer prompts, bro!
And conversely, some particularly quirky human paintings3 – for example, imagine that a painter sat at a canvas with paint and brush and painstakingly created a pixellated universe by hand, boxing her brushstrokes into thousands of tiny squares – were flagged as AI by users. In fact, the more detailed the painting, the more realistic, the more technical skill required, the more likely the painting was thought to be inhuman.
The AI art most often mistaken for a human one was impressionistic: a cafe at night, the light glinting just so: even those of us with little art history knowledge have seen the splotchy water lilies and ballet dancers of the great impressionists.
Finally, the study asked people to choose their favorite painting, and indeed, most users selected an AI artwork. Therefore … therefore what, exactly?
I left the thread disgruntled: a “gotcha” was implied in all of this.
If I were asked to pick my favorite painting and it turned out to be AI, I’d be a sucker, right? The person who created the artwork with prompts would have really gotten one over on me! This is not normally part of having a favorite painting. Unless you really like paintings of Austrian landscapes and it turns out the one you like the most was Hitler’s.
What exactly does it mean to have a “favorite painting”?
In my early twenties I rented an apartment and got a cheap couch and had a giant, empty white wall above that couch. I couldn’t afford … well, much of anything. One day I was at Bed Bath and Beyond buying some sponges or whatnot, and saw a big framed “painting” of some twisty tree branches with blue blossoms and I knew it would look great over my couch. Moreover, it was $50, which, for once, I had to spend. Once the piece was installed, I was grateful to have it, and it really did improve the room. But I wouldn’t call it my “favorite painting” because who the hell painted it?
I assumed at the time that some individual regularly churned out over-the-couch type paintings and sold the rights to Bed Bath and Beyond for reprinting, knowing the prints would be installed above couches nationwide. I assumed that this person then used the money from Bed Bath and Beyond to fund the necessities of their life, including the freedom to also paint gay erotica or cyborg fights or something sufficiently avant garde that I, a Bed Bath and Beyond shopper, lack the vocabulary to describe it. That is, I did not expect that the person who painted the painting necessarily endorsed it. I thought they may well have painted it cynically. The painting may have been their day job. They may have felt about people like me much the way the band who created the Friends theme song felt about all the “fans” who checked out their band only because of “I’ll be therrrre for youuuu” and not for their other, undoubtedly more serious work. A lot of artists hate what they have to do to pay the bills.
Under this set of assumptions, the artwork above my couch was simply not eligible to be my “favorite painting.”
That said, many great paintings were very literally the artist’s day job. The Sistine Chapel was a good gig! The Mona Lisa was a portrait of some guy’s wife. I’m sure plenty of artists bitched about having to paint whatever the Medicis were into those days. But there were both Medicis and painters. There’s a “there” there. Events, however mundane, occurred in human history, and paintings exist as a result of them. If I unironically love something that a painter painted only to pay the bills, I haven’t been suckered. No one is yelling “gotcha.”
I have, at various times in my life, been considered a “writer.” I have done this professionally. People have sometimes enjoyed and benefitted from things I’ve written. When I was in high school, I received some kind of national writing award that came with a stack of identical little certificates that I was meant to physically include in my college application packets. At that time, I was also prolific: I applied to 125 scholarship competitions my junior year in high school, and had a very high win rate for scholarship competitions that required an essay. I assume there were few entrants.
Between then and now, there have been several eras of writing. Blogs were invented, and there was a boom of blog-to-book deals4. Writers like Glennon Doyle, for example, were able to jump from blog to bestselling memoir.
Then, social media more or less killed blogging. Blogs that were “useful” sometimes persevered, albeit with a “lead magnet” (get the downloadable PDF!) on every page and affiliate links everywhere.
Then, because we live in a world optimized for the stupid, lazy, and hurried, Google decided it would prioritize webpages that answered users’ questions directly.
There was an e-commerce writer I used to follow in the mid-2010s – I bookmarked his blog posts, and found his advice useful enough that I started attending his annual conference (goodness, this sounds familiar).
Eventually, he started a podcast, which I also listened to. In one episode, he talked about how his Google traffic had tanked because his posts did not directly answer questions that people were searching for. One of the things he had enjoyed about writing was sharing his personal experiences, but if he tried to insert an anecdote into a post, Google no longer knew what the page was about and would deprioritize the content.
In response, he stopped writing. He hired foreign writers — people who were not e-commerce experts — to churn out articles on the precise topics people were searching. His blog now hosted articles with titles like “How to Start E-Commerce Biz With No Money” and “Shopify Alternatives Cheapest.”
Today, there is no need for the foreign writers at all. AI can do this for you!
In 2022 or so, I got an ad for software that promised to make a 40 page website for you on any topic at the click of a button. AI had not yet hit the mainstream, and I was astonished at this claim. The website offering this software linked to a sample website, an acceptable-enough 40 page website about tourism to Punta Cana, Dominican Republic. A video showed how this entire website, complete with images and internal links, was generated in a few minutes with a short prompt.
But why? What is it for? Who would want an instant 40 page website about Punta Cana? I mean, I would also be impressed if a machine instantly extruded a ten foot tall sculpture of an object of my choice out of bat guano, but it doesn’t mean I want one.
The answer seemed to be that you would host this website – and presumably many others like it – for the purpose of SEO backlinks. That is, you would crap up the whole internet for the purpose of linking back to your actual website from all your fake websites, thus tricking Google into thinking that many websites link to you because they find your content valuable.
Machines make content in bulk for the purpose of helping us trick other machines into showing humans some other thing we want to show them. What could go wrong?
Throughout the last twelve years or so, I have not been prolific at all. I have been running a business in a challenging economy; I have felt that it is not proper to write about petty concerns in the face of fascism; I have seen writers I know exhaust themselves of every possible personal experience and insight but have to keep pulling from that dry well to pay the bills. I have felt there is virtue in shutting up unless you really, really have something to say. Although perhaps that attitude now feels quaint as the internet bloats with AI-generated slop.
During the pandemic, I — once a philosophy major — spent perhaps thirty hours reading Seneca, choosing some of his best letters, rewriting them in a simple, punchy style, and adding copious swear words. I published these on this blog as “Letters to a Motherf*cking Stoic.” It was certainly unexpected, a little dark, perhaps cathartic.
Of course, most of the work was Seneca’s. But I, a human, took great care to select letters I thought contained advice that spoke to our trying times, and that would be funny if I added the word “motherf*cking” enough times. It matters just where you insert the word “motherf*cking,” even in a sentence that has many grammatical opportunities to do so. If you read these posts, you feel – I hope – as though you are having a moment with me, a human.
You may know that children in the US are, in many schools, no longer permitted to bring nuts to school because of allergies. Once, being charged with making my child’s school lunches, I bought a soy nut butter substitute that had thousands of good reviews. I tested it: it looked, smelled, and tasted remarkably like peanut butter. Impressive! And as soon as I had eaten one spoonful of the soy nut butter, I had no desire to ever eat another again.
It was a very convincing substitute, one that made me feel like I was fading away towards a painless death.
You see here that I have made a metaphor, and Google now punishes metaphors. Google wants to know what this article is about — AI, surely — and what questions it will answer clearly and directly (absolutely none). Now that I have mentioned peanut butter substitutes, the process of Google cataloging my content by its machine-perceived usefulness is well and truly obliterated.
The problems with asking human writers to provide clear and direct answers to user questions are threefold:
- This is literally the job of AI, and your attempts to jump in on it are only going to help train the AI.
- Answering questions people already knew they had is not that interesting.
- And finally, the most commonly googled questions are frequently asked by idiots, or by people passing the time on a phone while using the toilet; full of sound and fury, signifying nothing.
We have entered an era of “enshittification,” a term popularized by Cory Doctorow in 20225. The enshittification has since intensified.
I once, after a trip to visit my brother during which I began watching his favorite show, googled the question “Do Charlie and the waitress ever get together?” Google recommended a page that informed me that they do, but it doesn’t last long; the rest of the page was a lengthy FAQ about Charlie – going to space, fighting crime, marrying several different women, etc. It was in fact an FAQ about every character named Charlie from every TV show ever. According to Google, this is an excellent webpage. It directly answers everyone’s questions and is easy for AI to scrape.
A human cannot compete with this, and indeed should not.
I run this blog, GetBullish, which hosts articles about “careers and entrepreneurship from a feminist perspective” that I wrote between 2010 and 20146. Perhaps we could call these “classic articles” or “legacy content.”
I also run an e-commerce store – 5,000 fun and funny gift items in a little warehouse in Brooklyn – where it has been a struggle to keep the proverbial lights on since the 2021 Apple iOS update that tanked everybody’s advertising – and since everybody became perpetually worried and depressed, probably.
It seems that the “obvious” thing to do would be to publish little articles on the blog on topics like “How to Use a Cover Letter to Land a Job in 2025.” This is not a topic in which I am particularly an expert, so it may well be that an AI could do this better than I could. In theory, people would Google this topic, some of them would land on my website’s version of this article, and a subset of those people would click over to my store to buy something.
It’s not the worst business model, but like the narrator of Notes from the Underground:
I am a sick man. . . . I am a spiteful man. An unattractive man. I think that my liver hurts. But actually, I don’t know a damn thing about my illness. I am not even sure what it is that hurts. I am not in treatment and never have been, although I respect both medicine and doctors. Besides, I am superstitious in the extreme; well, at least to the extent of respecting medicine. (I am sufficiently educated not to be superstitious, but I am.) No, sir, I refuse to see a doctor simply out of spite. Now, that is something that you probably will fail to understand. Well, I understand it. Naturally, I will not be able to explain to you precisely whom I will injure in this instance by my spite. I know perfectly well that I am certainly not giving the doctors a “dirty deal” by not seeking treatment. I know better than anyone that I will only harm myself by this, and no one else. And yet, if I don’t seek a cure, it is out of spite. My liver hurts? Good, let it hurt still more!
I just won’t, because I won’t.
The more AI generated content you come across, the more you discover that you already know what it’s going to say, although I sometimes think this content is probably very helpful for, for example, someone who was raised in a bad home and never had anyone around to offer common sense advice and middle-of-the-road opinions.
A simple version of Hegel’s idea of thesis, antithesis, and synthesis is that when you combine an idea with its opposite, it becomes more true: the synthesis resolves the conflict between the thesis and the antithesis, preserving the truth of the thesis while incorporating insight from the antithesis. Then you keep doing it, repeatedly, by using the synthesis as your new thesis and transforming it through combination with a new antithesis: boom, the Hegelian dialectic!
In actual practice, if you do this enough, you sound like ChatGPT.
Something is true! But sometimes it is not true. Therefore, a thing is often true, but sometimes not. Every assertion becomes watered down, every exception elevated as a warning or caveat, until even the mildest assertion is equivocated upon. Everything is moderately correct except when it isn’t, which is often; you should always consider all the considerations, and the caveats to those considerations, and the considerations to those caveats.
The only human response is to write something that AI couldn’t write – or as that becomes increasingly impossible, to write something AI couldn’t write without lying, as an AI has no personal experiences and no convictions.
Even that will soon become difficult, as we will have fewer people with the skills or the will to read such things.
What is the point of writing anything when AI could do it worse?
We are in a liminal time in which a person can do what I am doing now: write something that an AI would not be likely to write, that does not already exist on the internet, that is tied to one human person who aggressively puts themself forward like a person in a shipwreck keeping their head above water hoping a passing vessel will spot them.
Joan Didion once said that writing is a hostile act7. Asked to clarify, she said “It’s hostile in that you’re trying to make somebody see something the way you see it, trying to impose your idea, your picture. It’s hostile to try to wrench around someone else’s mind that way.”
Perhaps this is the thing AI cannot do. It lacks hostility. That, my friend, you will find here in great quantity! I invite you to join me on a hostile journey, as there is much to be hostile to, and hostility in some contexts is a virtue.
Endnotes
- https://x.com/cremieuxrecueil/status/1859356055760752978 ↩︎
- https://www.astralcodexten.com/p/how-did-you-do-on-the-ai-art-turing ↩︎
- “Human paintings” is one of those phrases that was not needed until some new alternative had a public light shine on it. Once – perhaps the early 1990s – some number of straight people objected to the word “heterosexual” because they felt they should just be called “normal.” Today, cis people complain about the word “cis,” a word we could perhaps compare to “neurotypical” or “allistic” (both usually meaning “not autistic”), words not in common usage fifteen years ago, but also words that no one expects you to apply to yourself outside specific contexts. No one is asking cis people to call themselves cis on a daily basis; there are very few situations in which the average cis person needs to specify that they are not transgender. One more example, neither here nor there: I was once curious whether prosthetic racing legs, such as those used by Oscar Pistorius before he went to prison for murdering his girlfriend, were close to becoming better than actual legs for running. The study I read concluded that no, while racing legs have greatly improved, they are not likely to outperform “biological legs” anytime soon. I marveled at the phrase. We had no need for the phrase “biological legs” until we had synthetic legs. And we had no need to refer to “human-made paintings” until we got inhuman ones. ↩︎
- Between about 2005 and 2012, a number of friends had books published (several of which I am quoted in) that ten years later would’ve just been a funny Tumblr and today would be … well, maybe just a Reddit sub full of funny stories that don’t really need to be curated by anyone. ↩︎
- https://pluralistic.net/2022/11/28/enshittification/#relentless-payola ↩︎
- During the Obama era I could not possibly have imagined the far right resurgence in which we find ourselves. Personally, my first inkling at how bad it was about to get was during the 2012 election, when a teenager – openly, with his full name – posted on Twitter “Make the White House white again.” This was shocking in 2012. Today it is the entire zeitgeist on X, which resembles the sort of site that used to have to be hosted on an offshore platform immune from international law. ↩︎
- https://www.theparisreview.org/interviews/3439/the-art-of-fiction-no-71-joan-didion ↩︎
Want to support GetBullish? Send a gift card to a friend or buy one for yourself that you can spend later on your holiday gift shopping.
