Drawing conclusions from AI in Art

If we writers and creatives had a dollar for every time we heard ChatGPT or some other AI was going to take our jobs, we’d all be retired and lazing our days away on private islands by now.

In the writing space, the effect of Large Language Models, such as ChatGPT or Google Bard (now Gemini) was felt almost overnight.

Evergreen content (e.g., “Why does the sun rise in the East” type posts), filler copy, and other bits and pieces were and are being outsourced to Generative AI. That meant a lot of the dependable, “low hanging fruit” work for writers and marketers has all but dried up. Graphic designers too lamented the rise of automated design platforms such as Canva, Crello, or Snappa.

But what about the illustrators? The men and women who weave wondrous pictures using their imagination and a few brushes or pencils?

Illustrator Matty Taylor, aka KARMAVOID!, has been illustrating as long as he’s known how to hold a pen. He’s been told off for doodling in official document margins, on large expanses of butchers’ paper at supermarket delis, and of course, during boring maths lessons.

Honing his craft when bosses weren’t looking, he seized an opportunity to go full time in early 2020. He draws gnarly occult-style creatures and hellscapes for metal bands, lush tripped-out scenes for comic books, and was a background artist on Adult Swim’s YOLO, a homegrown surrealist adult animation. He’s now art director of the currently in-production Vampire Surf Cult, which is as awesome as it sounds.

What was Matty’s reaction to Stable Diffusion tech, the AI technique underpinning DALL-E, Midjourney, or Adobe Firefly’s ability to render images in seconds?

“Initially, it looked pretty scary,” he says, sipping a Vietnamese coffee. We’re at the appropriately named Yolo Café in Melbourne’s South-East enclave of Springvale. With tacky woodgrain wallpaper wrapped around the walls, only a human could design something this kitsch.

“You can pump out art with just a couple of prompts, but it’s completely devoid of life and enthusiasm. As someone who draws for bands, you see so many of them using it for album covers and things. That sucks because inherently it’s sort of against what those bands should be doing, at least I think, anyway. They’re artists wanting people to support them so they should technically be supporting artists in kind. But some don’t, and that’s just the way it is.”

Stable Diffusion is a Generative AI that’s trained on millions, perhaps billions, of images. After initial training by a human, an automated process categorises pictures, tags them semantically, and feeds the results into the model. From literal noise, an AI takes words in a text prompt to “extrude” out an image based on its training data.

What’s scary for Matty is that these models can be trained on artists’ own unique style and (theoretically) pump out similar looking pictures which eliminates the human element. Or does it?

“I’ve done it to myself before,” Matty says. “I wanted to see what it was like. I’ve also asked what it thinks about my art. I’ve asked it to rephrase it in a negative light, which is a very interesting thought experiment. It can be a real scathing bitch,” he laughs. “But these are all fun things. I think they’re there to be used, so why not use it? But interestingly enough, putting my own art in as a prompt to something like Adobe Firefly… It doesn’t capture the essence of it. Even visually, or the overall vibe.

“Art’s a big vibe thing. I think within about two or three seconds, you can tell if you’re vibing with a piece of art. That’s almost impossible to replicate with AI.”

Anyone who’s used Stable Diffusion knows how difficult it is to make minor changes. With overlapping noise and recursion, even a near-perfect result may tip over into nightmare fuel trying to change a single ‘inconsequential’ element.

“People are worried that in say, conceptual art, that space is being taken over by just pumping in prompts into an algorithm that’s shooting out all these different things,” he says. “The problem then with that is that you’ve got somebody like a director who will come in and want you to change this image that you’ve got from the ether, and you can’t, because every time you pump in those same prompts or variate them, it’ll change the entire image. You can’t work like that under deadlines. It’s too nuanced, and you need a person to get it right.”

At the end of the day, art’s value is in the eye of the beholder. One person’s AI masterpiece is another person’s utter AI slop.

“There’s no depth, there’s no nuance to some of the AI stuff, even if it can be very funny, like the Harry Potter as Balenciaga thing.

“For example, let’s talk about Pokémon cards. My 11-year-old son treats them like gold. I use them as a bookmark. You know what I mean? It’s all about shades of value. People will still value original art.”

Join the soloist movement. Whether you are new to Flying Solo or looking to grow your business, our membership options will help you attract more leads, grow your network and sharpen your business skills.  Sign up to our newsletter to get the latest news and advice straight to your inbox.