The Problem With AI Art: Revisited
Two years on, how different is the AI Art landscape looking?
Two years ago I published an article called The Problem With AI Art. Since then, a few things have happened.
That article was published in Data Driven Fiction, a publication that is sadly, no more. As a result, it has now been added to The RGB Chronicles.
At the time, I knew that AI Art was going to cause some disruption. However, the shock of the new has begun to fade, so now feels like a good time to dust off some of my initial musings on AI Art to see how my feelings have evolved. Here is the original article:
At the time, I likened the process of writing an AI prompt to ordering a sandwich in a cafe.
The process it most closely resembles is ordering a sandwich. I’ll have salami, tomato, cheese and pesto on wholemeal seeded sourdough. Like the sandwich, it’s hard to imagine that its ability to nourish extends beyond a few hours.
This remains largely true, yet many of the AI Artists have overcome this by making the same sandwich over and over again. This might suggest repetition, but what it more closely resembles is the development of a style.
The development of this style also creates an impression that no single work is a destination as such, merely a step on the artist’s journey, something I sensed the instant gratification of AI Art might destroy.
There are two key consequences of this.
Firstly, no single work by an artist is memorable in its own right. This is a blessing in some respects. The recognisable veneer of AI Art is subjugated by the overall trajectory of the artist’s conceptions.
Secondly, the development of this style requires the creation of a mind blowing quantity of work. I recently saw an artist on X claim that they had created over 500,000 works. That seems impossible, so let’s have a quick look at the numbers.
The image in the title took about fifteen seconds to generate plus another fifteen seconds to type in the prompt. Most AI generators will produce four variations for each prompt. That’s eight images a minute or 480 images per hour.
DALL-E was initially released on 5th January 2021. So roughly 1,200 days have passed since it became available.
1,200 x 480 = 576,000 images. In less than an hour a day for the last three years or so, reaching the 500,000 mark of images created would be entirely feasible.
In the realm of artistic creation, the only thing that can come close to achieving such quantities are frames in a film. The standard cinema frame rate is 24 frames per second (FPS) so a two hour movie has 172,800 frames, a number approaching the quantity of images that an AI Artist may have created.
Coincidentally, those 172,800 frames correspond to about a year of effort on the part of an AI artist. I believe the coincidence points towards how AI Artists might develop their work into something more substantial. Video.
To date, the most famous and recognisable AI artwork is Théâtre D’opéra Spatial by Matthew Allen. That fame and recognition comes mostly from persistent repetition in the media. It’s a banal slab of kitsch that recalls Jack Vettriano more than Michelangelo.
Ultimately, much of the AI Art I see struggles to be compelling as individual images, each image working best when they are a component in an overall scheme.
Some of the most vocal criticism I’ve seen of AI Art comes from illustrators who see a most direct threat. I expect these concerns will dissipate over time as the recognisable surface of AI Art will become even less compelling as individual images and its suitability for paid commercial illustration will fade. Like the boy screaming into a microphone on Unsplash, its utility will be reserved for unpaid illustration work, such as the image used above. I can’t imagine there is a single writer on Medium who pays an artist to illustrate their work.
Conclusion
Two years on, I feel the problems I originally identified remain. Some artists have overcome this to an extent, yet this has thrown up another set of problems and challenges.
Finally, I’d like to touch on a few artists that have got my attention and their work forms the basis for some of the observations I’ve made here. From my perspective, X is the best platform for keeping up to date with their ongoing creations.
WARNING: ADULT CONTENT, NSFW.
Vnderworld
Navigating a hellish, nightmare world that might keep Hieronymus Bosch from sleep, the worlds are full of demons, both personal and imagined. The intersection of these demons is best demonstrated by this video, #12 — Demon’s in my Whiskey Bottle, a compelling rumination on the torment of alcoholism, each frame operating as a scene in strange, non-linear movie. The clip is at the bottom of the embed.
Vnderworld has also achieved recent acclaim for his work with Peter Gabriel.
Splarg
Is splarg a verb yet? If not, it will be soon. Splarg’s work is nauseating in the most literal sense. His X timeline is a bombastic assault on both the senses and the stomach, and like ouzo and coke, I wouldn’t recommend consuming too much in a single sitting. His video, The Misery of Life and Death, further demonstrates the power of AI images as scenes in movies. The added context invests each image with more of a story. Try not to splarg.
カタクリコ (Com Maruko)
While Com Maruko hasn’t put together any video than I’m aware of, the entire feed feels like scenes from a bizarre retro TV show. The only thing I can think of that comes close is Patrick McGoohan’s The Prisoner. Like The Prisoner, the action takes place in a remote location replete with laboratories, talk show sets and plasticly, picturesque countrysides. The style is instantly recognisable, not least for recurring themes of sushi, fish, cats, strange rabbit costumes, a malevolent military, nuns, squid, retro sitcom and drama characters and caviar. Lots of caviar.
Guest appearances from Snow White, Harry Potter and a host of superheroes neatly echoes the role of Number 2 in The Prisoner.
These profiles skim the surface of some of the interesting artists that have begun to emerge in the AI art arena. As I revisit more of my AI articles, I’ll try to include a trio of relevant artists with each.
In the meantime, if you’re on X, here are a few more I’d recommend following. Jump in and have a look. As I’ve noted above, it isn’t so much about the individual works, but how those works come together to create a meaningful whole, so following along with the journey is part of the fun.
D The Designer • Lame Times • Particular Uno • Lynn Cole • Poopy Pants • Hogan Torah (yes, Mr Autofill)
There’s many more, but this is a good start. Please enjoy.
Finally, a shout out to evaARTology. Thanks for all the work you put into Data Driven Fiction. You can follow her on X here:
Hi, I’m The RGB Designer, AKA Robert Gowty. I’m a writer and digital artist. You can read more about journey here:
