2022 changed the art world forever.
In the blink of an eye, everything was thrown into chaos. “AI can make WHAT?!” Everyone choked in horror. Mass panic tore through every art community in an instant. It only got worse over the next year, when tools such as Midjourney showcased their system’s improvement. The results were jaw-dropping.
What’s worse was that there was a myriad of generative AI such as Dall-E or Stable Diffusion. (Today, there are more than ever before.) Artists feared for their jobs and rightfully so. A skill that takes years to hone is seemingly replicated in mere seconds. It didn’t help that tech bros demeaned artists, telling them that their passion and years of study were suddenly worthless.
As hysteria rocked seemingly every artist from beginners to 40-year vets, the optimists and realists tried to provide the community with some comfort. They showed how these systems functioned, and how AI could help artists in the creative process. Every artist needs references, but it can take hours to find them. AI, however, can provide custom references. Besides, the optimists and realists rationalized, AI will never be superior, because it is not human. It does not have memories, feelings, passions, or fears like humans do. Human art is genuine expression while AI only follows commands.
The panic slowly subsided. Many artists began using AI for their paintings. Fear remains, however, but not for job stability. Instead, many artists fear for the safety of their own intellectual property.
AI images are trained on databases of countless images1. To heavy simplify things, each image in the database has a series of tags as identifiers, such as “yellow grass,” “gold wheat,” “Vincent van Gogh style,” “blue sky,” etc. When users enter a prompt, the AI searches through its database to find works with similar keywords and stitches then stitches several together to generate an image that matches the prompt as closely as possible.
This raises ethical concerns. Many modern artists whose art is not in the public domain have had their art added to generative AI databases without their permission by developers who don’t think to ask2. One major controversy in the art community surrounded Canadian artist Sam Yang, a popular figure in digital art circles. His portfolio of over 300 unique paintings and sketches was illegally used to create an AI that generated images in his signature style. This is blatant copyright infringement. Tech companies and independent developers are shamelessly committing IP theft without so much as a second thought.
What do we do? How should developers be held accountable? How can AI and artists possibly coexist?
The answer is simple: AI developers must respect copyright and trademark law. IP law is a system that is frequently updated and adjusted to protect the rights of those who dare to create.
Of course, many of these developers are anti-copyright communist twats who don’t care about ownership until someone steals their property. However, lawsuits are becoming increasingly common, such as Getty Images’ lawsuit against Stability AI. AI developers must realize that the world doesn’t belong to them and them alone.
Firstly, if a development team wants to use a specific artist’s work in their database, then they must receive permission from the artist. And—if the artist is smart—the team would be required to pay royalties to the artist for continuous use of the art in the AI’s database.
This would open the door to a new revenue stream for artists. By being a real “AI artist”, a creative could make art solely with the intention of training AI. Not only that, but the companies behind these tools would avoid possible class-action lawsuits.
The only artworks that companies and independent developers would not be required to ask permission for are those within the public domain or with open licenses (e.g., Creative Commons)…You know, just like everyone else.
Secondly, AI companies should make it a common feature to display which particular artworks are used when generating an image. (Last time I checked, there are no generators that do this.) This would provide artists—both living and dead—with the credit they deserve.
Additionally, it would allow inquisitive users to further understand how the generator works. While they wouldn’t be viewing the entire art database or the AI’s neural networks, it would at least allow the average layman to peel back a layer or two. Casual users would gain a deeper appreciation for the AI’s capabilities, whether the tool scraped together elements from two artworks or two hundred.
And, again, lawsuits could be avoided. The choice is up to AI companies and developers. At the end of the day, they can do whatever they choose.
But, hey, I’m not the one facing legal action.
Some people believe AI generators crawl the internet for images. This is not true. Humans feed the AI datasets.
If you or someone you know is an artist and want to protect your art from theft, I’d recommend using Glaze, a free tool uses AI to disguise your art’s true appearance from generative AI.
UPDATES: Stability.AI is being sued for previous versions of Stable Diffusion. Now, it regularly uses stock images for its training. Dall-E also has retrained its datasets to exclude human artists.
Yes, those things would make AI art ethical. However, the companies training the AI models aren't ethical. They just see dollar signs.
When I have the AI create art, I try to uncheck any box that has an artists name. I try to make everything original.
I found this post a little late but it instantly reminded me of the impact on music by technology such as Napster and digital recording, which ushered in sampling and drum machines.
Copyright enforcement resulted in closing file sharing companies and Apple created the first streaming model that gave pennies to the artists. Samples became highly regulated.
Drum machines reduced session and live gigs for drummers.
All these and many inventions since, totally changed what it meant to be a musician. Now, people can master technology and create music instead of mastering music... at least live musical performance.
Perhaps studying how technology changes the music industry could provide ideas to what may happen to visual art.