NOT A FAN OF AI-GENERATED ART
AI-generated art sucks, and not in the way you think. Oh, way to go — the first sentence is as clickbait-ese as the title. What kind of lazy, amateurish shit is this horseshit blog?
Look, seemingly out of nowhere, AI art applications produce way better art than some humans who work professionally in design and art.
And one could easily predict human artists losing jobs or money to a technology that whips together finished compositions in the time it takes to cycle a few coins into a parking meter.
Is there a future where human artists need to underbid technology companies that have perfected fast and easy art? Will Disney eventually reduce staff for computers that spit out new animated films in record time?
In this anti-utopia, are medical companies 3D printing human bodies for medical research? Will that remove the need for grim checkboxes like “donate your body to science?”
Man, that got dark… Quick type “Monsters University” into an AI-generated art site.
It may be time to begin the problematic morality versus technology questions before us. Steak is being 3D-printed, for god’s sake. Will there be a time when organ donation checkboxes are removed from driver’s licenses because we can 3D-print them?
Will companies loan out these organs charging monthly rentals? Maybe sponsorship content plays through the host’s vision? Personal commercials that play like: “this part of your life is brought to you by Depends; get back into life with Depends!”
Gawd, I love dystopia. It can have bits of “well maybe” while wandering off into the unapologetically absurd. But that’s not why we’re here.
WHAT IS ART?
Art can now compete with the tree that falls in the forest, thought experiment. Is art, art if no one (at all) sees it?
Or we could start with an easier one… essentially defining art and evaluating its worth based on who makes it.
By the way, I will not stand on mount artist and defend the entirety of human creativity from artificial intelligence. There will be plenty of heavy-weight brains on both sides to battle that shit out (and again, that’s not why we’re here).
I’m on an adjacent career path with these folks; if the art itself can be replaced, so can the guy who edits it. So, I (obviously) weigh artists’ opinions more.
Objectivity flits around their work like misery sprites. They could create what they would define as a great piece and receive blank stares and ambivalence in return. Technology will never experience the truth that cuts like a dagger through the thoracic spine; it’ll just spit out something else.
And art doesn’t exactly pay well — the earth is rich with the bones of the most talented artists who died poor and unknown. Trodden their entire lives to only become earth-shatteringly famous long after they gave up brushes for bones.
And whose work now hangs in the halls of the most powerful.
Catastrophe can be rewarded in art. Especially after you’re in the ground and someone is trying to offload that lighthouse painting from 1984. Nothing moves art like a tragedy.
AI-GENERATED ART SUCKS BECAUSE IT WON’T EVER FEEL REJECTION
AI could eventually pose a genuine hurdle toward artists being paid for their labor. And that’s a big deal when your skill to most other humans should be free of charge anyway (ask your artist friends if they have ever been propositioned for free work).
There is a maxim in the art community of avoiding work that pays in “exposure” or creating something for someone who will then use it for free. And that somehow benefits the artist.
AI-generated art presents a place for those exposure-employers to run wild. And it may get ugly (at least early on). Be prepared to encounter flyers for car washes that look like fucking Gone with the Wind movie posters (more on this in a second).
Future artists could see years of techniques and learning used for exposure against their will. And not exposure for the artist but exposure for the AI.
“These systems are trained on existing material, often using content pulled from the internet,” AI-generated art’s bedrock is potentially all human artists with a digital footprint; ever — AI could foreseeably gank styles from centuries of designers.
One artist had their work stolen as they were working on it and “finished” by AI before the artist could post the finished piece.
LOOK HOW FAR WE’VE COME
That AI-Generated art exists and looks as good as it does at all speaks to tremendous technological accomplishments. Similar advancements brought about design tools, social media, personal broadcasting et al., that artists use to become luminaries in their living years. Or at least known and rewarded for their content in real-time.
That technology can give and take away human advantage at almost the same time…
As mentioned, there is a market of customers who want fast and cheap art. And AI-Generated art in its present form will fill that want. Even if it doesn’t give the clients precisely what they were hoping for.
All you need is a few keywords, and AI will create some sensational shit in twenty seconds or so. AI also makes amazing shit with next to no keywords (see above).
Hopefully, putting to rest any thinking that filling out a text field and fingering the return key has been an artistic journey.
That’s not to say there aren’t options that bring the user a bit closer to the creation of the piece. There are inputs for adding photos from your device and assigning a strength of influence to adhere to.
I took a live photo of the sky (I’m a fan of clouds), entered a prompt, and picked an “art style” — a quick progress bar initiates, then a flash, and then art I had mostly shit to do with arrived.
If anything, I’m a witness to a digital snowflake. Something beautiful that melts away immediately unless special care is taken to catalog it. A digital art equivalent to Wilson Bentley, perhaps.
A BUMMER TO REFRESH
It’s disheartening to flush results of AI-generated art for any reason and generate anew — as if the prior work meant nothing.
There’s a scene from On the Beach where technology is discovered blindly marching onward after an apocalypse that left populations in the United States, and most everywhere, dead. The book and movie differ slightly on how a call button is clicked to create what the few survivors south of Sydney hope is morse code. I prefer the simple Rube Goldberg explanation in the movie.
And it strikes me that AI could be forced to play out this demoralizing theme one day. The weight of a fallen object on a keyboard that forces a constant refresh — cycling accidentally through years of beautiful AI-generated art. Shown to no one. Because we are all dead.
Ugh, dark again.
Ok, let’s imagine everyone is still alive. There’s this artist from Uruguay on a business trip, and she left her cat. And the cat is sleeping on a keyboard that forces hours of AI-generated art shown only to some asshole cat. Staring up at results in the classic cat expression of confusion/ malice.
LINKEDIN MARKETING FOLKS SEEM TO DIG IT
I’ve seen the posts on LinkedIn, the “look what I made” posts, and rhapsodizing their “new tool.”
I imagine easy smiles on folks’ faces as they experiment with long-dead artist names in text prompts. They save a slew to post in marketing groups. Clacking out confident endorsements. Lift coffee to their face as LinkedIn emoji applause rolls in. The consensus is hurrah technology.
And then all the LinkedIn marketing thought-leaders brush their teeth and go to bed. They don’t even finish out the day — tuck themselves back into bed, being sure to set the alarm for 23 hours later.
It was just that one thought needing to be overwhelmingly loved. And tomorrow there will be a new thing.
Maybe I’m being unfair, but what the hell are y’all talking about? Has anyone seen the AI-Generated Art they are posting for timeline accolades?
The shit is goddam terrifying. Visually assaulting scenes that look conjured from the same place nightmares are born.
DO YOU REALLY WANT THIS?
“You don’t tug on superman’s cape,” and I don’t fucking carbonate chicken soup just because I own a SodaStream.
No technology should take me seriously if I ask for “Wade Boggs” and “cabbage” or “King Kong” and “Canterbury Tales.” I shouldn’t task anything with trying to visualize “Jesus” and “works at Starbucks.”
But now I can.
And look, I’m respectful of religion (at the very least, I have rules for avoiding money, politics, and belief in my writing). My intentions are not for shock or blasphemy — it’s just that you can type in the most absurd shit without a human voice being like, “Starbuck Jesus… you sure?”
Interesting queries that would be too pricey to explore with human artists can now be commissioned — like “Margaret Thatcher” and “baseball card;” or “Calvin Coolidge,” “hip-hop DJ.”
You could create new works featuring the most harmful human beings or villainous ideologies the world has ever produced. If there is visual evidence for the AI to learn from online, I bet AI could pair it quite easily with something absurd like a Christmas card.
Most human artists aren’t working with you if you’re hoping to mash up the most infamous dictators with Santa Claus.
Not the AI; the AI delivers you your visually engaging nightmare shit — zero arguments.
AI CAN MAKE YOUR HERO SUPER BUT WITHOUT A PUNCH BEING THROWN
You may spot AI-Creations displayed at your uncle’s house this holiday season. Something he discovered being sold at the county fair that featured his favorite politician in the Iron-Man suit.
One thing AI does well is simple mash-ups. I tried “Kathy Hochul” plus “Superwoman.” And the AI insisted on kicking out some fucking epic shit.
But you cannot use that generated awesomeness again in a modified scenario. From what I’ve seen, you cannot place your super Kathy Hochul into a scene punching Justin Trudeau and then flying downtown to stop a bank robbery.
It takes a lot of work to direct characters in the composition to do anything specific. And after countless refreshes, you will see breathtaking examples that are far from what you were hoping for.
Say you wanted to create a Godzilla monster destroying a city while drinking its morning coffee. The AI will return with an impressive monster wreaking havoc and like a random floating coffee mug.
If I commissioned an artist to create a piece showing a “woman robbing a bank,” I could be as specific as I wanted with the artist.
The robber could be a diminutive cybernetic lady with an eyepatch and a large cigar protruding from her lips. She holds her Scottish Claymore with her third and fourth hands as clown-nuns take cover in the background. Basically, I’d be hoping Rob Liefeld was open for commissions.
This isn’t something I could direct the AI to do.
I couldn’t even get the AI to visualize “woman robbing bank.” Sometimes I received a woman holding a weapon but no bank in sight, then a fashion illustration with pillars.
Never what I assumed; a woman facing someone in a bank, maybe with their hands in the air.
SMUTTY ARTISTS NEEDN’T WORRY (YET)
Further, suppose you’re hoping to generate naughty stuff. An image of a tryst that features Edgar Allen Poe and the lady from the Golden Girls, Big Bird making out with Cleopatra, or a massively naked Statue of Liberty.
AI generators that can pull off bankrupt-desire commission art fast would be something I see doing well. Customers too timid about broaching the depths of their depravity with an artist who might call the cops.
Yet the image generators I tried couldn’t achieve a nude body or anything remotely NSFW. And I bet that’s by design. (BTW, I wasn’t the only person wondering).
So the subset of DeviantArt-ish folks that scratch out a living catering to thirsty clients’ demands seemingly has nothing to worry about.
Speaking of DeviantArt, the site has dipped its toe into AI-generated art, and it’s not going well for the human users there.
AI-GENERATED ART SUCKS FOR A LOT OF REASONS
Adapt, it’s something humans do well. I see that being the main argument for artists unhappy with the existence of AI-generated art.
And if you’re ok with that argument, just remember it when computers run entire restaurants. And your fave gyro spot has to shutter because it can’t compete with an AI chef that has learned from centuries of grandma’s best recipes.
Ahh, that sweet, dark dystopia.
Look, if LinkedIn marketing folks in your timeline are going that gaga over it, best to fucking beware. Which (sorta) returns me to my premise, if you want something specific, ask a breathing artist to do it.
And please pay them.
Aside from the scores of viral AI-Generated art that suddenly popped up in my feeds, I mainly explored Dream by Wombo because its free. Most of the other sites want you to create an account or pay to play (and I’m not that curious on differences). Dream allows free use of its robust set of tools and creates some jaw-dropping shit.
It really does trigger internal “look what I did” energy — when I only spun a mildly customizable roulette wheel. And it is still a rush all the same. Entering ridiculous prompts and seeing silly art arrive in seconds.
Sometimes astonishingly beautiful silly art.
Although never without blemishes. Figures will be there, not so much interacting as melting into each other. And often with a plethora of unassigned limbs. Nightmare shit.
But nightmare shit with better design skill and color choice than I can often hope for in my professional work.
It’s humbling, unfair, and probably not going anywhere.
****Oh, and in case you haven’t spent enough time with me today, I started an ambience podcast! Go check it out!