Last week, a man named Keith Schofield did a tweet that went viral:
Good for Keith, I say. Unfortunately, there are two kinds of viral tweets: the good kind, which is when everyone shares your tweet and coos about how clever/cool you are, and the bad kind, which is when everyone shares your tweet so they can sneer at what a dipshit you are. Some people would argue that, much like “all press is good press,” there’s no such thing as “bad viral”; these people are sociopaths and are not to be trusted. Anyway, Keith went bad viral.
David Cronenberg never directed a movie called “Galaxy of Flesh”; in fact, nobody has ever directed a movie called “Galaxy of Flesh.” The “stills” aren’t even from a film, they’re images Keith created using an AI tool called Midjourney. The whole thing was a lie—including my use of the word “created” just now, which falsely implies that Keith contributed in some way to the finished product. Really he just typed in a prompt and Midjourney did the rest.
The reasons for the negative attention varied. Some people were mad that Keith didn’t add a disclaimer that “Galaxy of Flesh” is not a real movie. Other people were mad that Keith didn’t make it clear he’d created the images using Midjourney. Still others thought the images reduced David Cronenberg’s vast and diverse directorial work down to simple body-horror schlock and got mad about that, which is probably not the thing to be focusing on but is A Thing nonetheless. Mostly, people were mad because they felt duped.
At first glance, the images do look like they could be from a movie. It’s not unreasonable to see them and assume that David Cronenberg—who has dabbled in body-horror stuff in the past, to be fair—directed some low-budget sci-fi horror/thriller flick in the mid-80s. People were willing to accept that A) David Cronenberg made a movie called “Galaxy of Flesh” and B) that’s what it looked like, because it felt true enough. They added that information into their mental libraries of Things That Are Real, only to swiftly realize that none of it was. Hence, the anger.
But looking at those images, I don’t feel angry or embarrassed for having thought they were legitimate. I just feel intensely creeped out.
Upon closer inspection, you can start to spot things that aren’t quite right, like how the tubes in the images don’t seem to connect to anything, or the boneless quality of the hands, or how clothes just sort of…end:
The errors are sort of laughable once you know they’re there, but I didn’t notice any of them at first. And that freaks me the hell out.
Even when I zoomed in, my brain processed everything as a whole and decided it looked realistic enough to pass muster. It wasn’t until after I learned that they were AI-generated images that I started spotting all the things that were off about it.
In cases where the finished product is more visibly wonky, AI art often still looks realistic enough at a glance that your brain will just kinda glide right over it. For example, this set of photos, which make me laugh every single time:
If I look at this for just a split second…
…my brain files it under “Muscle Men Guarding JFK’s Limousine (Normal)” when it should be filed under “Noseless Mutants With Incorrect Abs and Severe Facial Burns Trying to Hide Cars Behind Their Butts.” And this is someone trying to lean into the weirdness of AI art!
Some people see those flaws and go “Aww, the computers are learning :)”; I see them and think oh fuck, they’re studying us. Everything about AI art feels off, and it provokes the same kind of evolutionary physical response I have when I see a close-up image of a spider’s eyes: my stomach knots up, I’m filled with a vague sense of dread, and I just want to put as much distance between us as I can.
I hope AI art never becomes widely adopted, and not just because it gives me the heebie-jeebies. There are a bevy of ethical concerns: AI programs are trained using the work of real, flesh-and-blood artists, almost always without those artists’ consent or knowledge. And because AI art is free—for now; rest assured that’ll change the second someone figures out how to monetize it—that means artists' livelihoods are threatened by programs they unknowingly helped create and for which they will receive no compensation or credit. AI art can also discourage people from learning how to create art for themselves; why spend years learning and perfecting your technique when a machine can crap out a reasonable facsimile in a few minutes?
AI art is also a microcosm of a larger dilemma. In the simplest terms, the value of AI is that it doesn’t need a user to tell it what to do every single time; with enough training and data, it can eventually learn what to do and, crucially when to do it. But the more we use it as a shortcut to simplify every aspect of our lives, the more it learns, and the less it needs our input. On a long-enough timeline, the balance of power is inverted: we become increasingly dependent upon it, while it relies on us less and less.
Crappy sci-fi movies about robot uprisings and such always make sure to remind audiences that although computers may get more advanced, mankind has something no robot could ever have: creativity. There’s always someone saying “A computer can’t paint you a picture of a sunset” or whatever, which is supposed to reassure us that a robot uprising isn’t really in the cards. But now a computer can paint you a picture of a sunset.
Art and creativity are part of how we express our humanity. Sure, it’s faster to have a computer make us a painting than it is to paint it ourselves, but that’s not art, it’s manufacturing. It’s convenience for its own sake. Art is about the process; you don’t paint so you can have a painting when you’re done, you paint because you want to create something.
We’ve been on this track for a while now, trying to find a technological solution for everything in our lives with no thought given to whether or not it’s even needed. As we speak, Midjourney is learning how to create photos from a party full of people who don’t exist. It’s probably just a matter of time until someone trains AI to generate memories of that party that never happened with those people who aren’t real; why bother showering and getting dressed and going out to a party when you can access the experience of having done all that right from your couch?
Maybe that’s a long way off, or maybe I’m being paranoid about the future implications of it all. But AI art isn’t a hypothetical, it’s here now, and it isn’t a solution to anything. It doesn’t enhance art, it cannibalizes it. It doesn’t make things easier or more efficient, it allows us to avoid the pursuits that make us human.
If we outsource those, what’s left to make life worth living at all?