It was only a few years ago we were talking in the magazine about generating faces with AI and, at the same time, pairing that with generated text. The zeitgeist of the moment was that it could be used to generate fake news on an industrial scale, or at the very least, some garbled stories for our idle amusement. Now AI is able to generate pieces of art, and it’s raised a thousand questions in the process. The most famous example of which is the piece above “Théâtre D’opéra Spatial”. It was generated using Midjourney by Jason Allen. He submitted it to a local art competition under the “digitally manipulated photography” category and won handily earning himself $300. He spent 80 hours of prompting and generating 900 iterations until he finally finished his series. It still takes work to get the pieces up to snuff. One of the pieces was missing a head, which he had to add in post while removing other artifacts. Now the question remains, is it art? Sure, why not? Is Jason Allen an artist? He seems to think so, he believes that this AI “is a tool, just like the paintbrush is a tool. Without the person, there is no creative force.”
He’s more right than he thinks. One aspect that AI wranglers cling too is the recipe, or the list of prompts they use to get their results. It’s a secret most will keep to themselves, and generally looks like a string of keywords you’d plug into Google. The awkward part is that this recipe usually contains a list of known artists that the AI zeroes in on to get a good result. Cribbing from artists has become an issue. Some popular artists are finding their own Google search results scummed up with AI knock-offs, diluting their brand.
So is it theft? One AI artist has already rushed out to copyright his comic book made of generated images, claiming it as an original work. However the data used to create the generator was pulled without permission from numerous art sites. This data was then monetised by the AI provider the AI artists use.
According to Wandb.ai Dance Diffusion, a music generating AI “is also built on datasets composed entirely of copyright-free and voluntarily provided music and audio samples. Because diffusion models are prone to memorisation and overfitting, releasing a model trained on copyrighted data could potentially result in legal issues. In honoring the intellectual property of artists while also complying to the best of their ability with the often strict copyright standards of the music industry, keeping any kind of copyrighted material out of training data was a must.”
Harmonai, Makers of Dance Diffusion, recognise the legal repercutions of what they’re doing. But this recognition didn’t extend to their art generator Stable Diffusion, perhaps because freelance artists are less likely to lawyer up as much as the music industry.
Perhaps the only thing Allen and his hyperbolic detractors are wrong about is art’s status. “Art is dead, dude. It’s over. A.I. won,” he told the NYT.
Art will never die because it isn’t a thing. It’s a personal journey we all take on our own terms.