After recording last week’s episode of On Taking Pictures, Bill sent me a link to a video that’s both fascinating and deeply disturbing, called “AI vs Artists: The Biggest Art Heist in History.” The video presents some of the grim facts around how images, including the 5.85 billion uncurated images in the LAION-5B dataset, are being illegally scraped and used to generate derivative work. The dataset was initially intended for research but has since been made available commercially and has been used to train AI models, including MidJourney and Stable Diffusion. While it does contain images from the public domain, it also contains millions of copyrighted images, as well as explicit content. As they say in the video, no consent was obtained, nor were artists given the opportunity to opt in or opt out—and this is really at the core of why so many artists whose work has been stolen are so upset. Also, AI models never forget, so once an image is used to train a model, it cannot be “unseen” and becomes a part of that model forever. To complicate things even further, Generative AI can be used to copy an artist’s style and create entire bodies of work based on it. Simply describe what you want and include something like “in the style of X” or upload one or two images by a specific artist as “style references,” and a few seconds later you have a collection of images that would have taken the actual artist days or even weeks to create. It’s not just the blatant theft of intellectual property that’s a problem with Generative AI, but also the potential damage to an artist’s reputation. Prior to the availability of tools like MidJourney, you had to have talent to rip off someone else’s work (or read this). Now, with a sentence or two, you can effectively copy an artist’s style—or at least get “close enough”—and create entirely new work that may get attributed to that artist, regardless of whether they had anything to do with it or if it even lives up to their standards. And Generative AI is not just available in one or two fringe applications. Adobe has built AI-based tools, including Generative AI, into many of their most popular apps, including Photoshop, Illustrator, Premiere, and After Effects.
So is Generative AI all bad? Objectively, I don’t think so, but answering it thoughtfully requires some nuance. For example, a photographer friend of mine has started to use the Generative Expand feature in Photoshop to change the aspect ratios of some of his photos. So while technically he’s using AI, it’s being used locally on his own work, rather than scraping the web to steal the work of others. And from what I’ve seen, the results are fantastic. I’ve actually done a few similar experiments using my own paintings, and my results, while lacking the detail of the original, have been interesting conceptually—in some cases, interesting enough to potentially send me in different directions on future paintings. But from the research that I’ve seen, that’s not how most people are using it. Most people are using AI—at least Generative AI—as a “one-stop shop” for making images, with the only human contribution being an often-nebulous text prompt.
Coincidentally, I recently got an email from an OTP listener with a link to a fascinating Juxtapoz article about a British artist named Wolfe von Lenkiewicz, who combines AI algorithms and traditional oil painting to create a body of work that feels incredibly modern while still conveying the artist’s connection to German Romanticism and French Neo-classicism. I spent some time looking at his work on Instagram and on his website, and honestly, I think it’s brilliant work that requires real skill and technique to pull off, regardless of how “good” or “interesting” the initial AI-generated ideas are. Removing the controversy over the source material for a minute, I can absolutely see the value in a workflow that uses Generative AI as a sketchbook or sort of a rapid prototyping tool, which is then used and incorporated by the artist into the final work. At least then, the process becomes a collaboration of sorts, and I think maybe it gets us closer to the more conventional analog equivalent of being inspired by the work of someone else.
I’m following an interesting discussion on Threads that was prompted by some recent comments Annie Leibovitz made around how she’s not worried about AI “at all.” Of course she isn’t. She doesn’t have to be because her name has become as much of a commodity as her work. One of the responses to her comments gets to what I think is at the heart of the AI issue for many creatives:
“The people worried about AI generated images are the ones that are making art on commission. Smaller artists or illustrators, who make/made money from (for example) getting hired for magazines, are seeing their livelihood taken away by generic AI stuff, because it’s cheaper.”
While I think that’s mostly true, Generative AI is also making its way into the more lucrative fine art world. There was recently a show at the Gagosian in LA featuring AI-generated work by filmmaker Bennett Miller, who spent several years interviewing some of the people who are driving—and in some cases, reshaping—what society may look like in the coming years on the back of AI.
Regardless of where you stand on any of this, the fact is the AI genie isn’t going back in the bottle. Companies and creators have each gotten a taste and they want more. Many creators see it as a way to make more work in a fraction of the time without possessing any “traditional” creative skills. On the other side, many companies see it as a way to pay less for work that’s “good enough.”
Personally, I think AI and artists can coexist, but it has to be an ongoing conversation rather than a one-off transaction—and I don’t think it can be sustained if it’s rooted in greed, laziness, or outright theft.
QUESTIONS
Are you using or have you experimented with Generative AI?
If so, how are you using it? If not, what’s keeping you from trying it?
Hit reply, leave a comment or email me at talkback@jefferysaddoris.com.
If you enjoyed this Iteration, I would love it if you would share it with a friend or two and if you’re not yet subscribed, maybe you could do that as well.
Thanks for reading.
That’s a hell-to-the-no from me. You already went into the ethical reasons (minus the enormous carbon footprint of the generators), so no reason reiterating that here. Though an additional note should be made about bias. Image and text generators largely rely on and reinforce race, class, nationalistic, and gender biases.
As to artists using it, sure, I suppose they can. But, honestly, why? If your photo isn’t wide enough, next time use a wide angle. And if you don’t know what might be in a wider version of your painting, well, get out the paint and paint it. Explore. Having the machine do it, won’t get you inside the work. I read an article by a writer who uses AI to generate scenic detail they don’t care to write about. But all that is process. Scenic detail — or lack thereof — speaks to a book’s point of view. What’s mentioned or not matters. And just filling a page with unintentional words is garbage.
There’s a passage in “We” by Zamayatin that describe’s a future “utopia” where people can compose several symphonies at a touch of a button. But symphonies hold no value for these people, who live under a surveillance state. Nor does love. This book was written in the 1930s, but predicts instant AI creation and the values of a society that uses it.
It’s not going back in the bottle. But I also don’t have to touch it. And I won’t.
Excellent post. Perhaps AI is like a gun -- you can kill someone, or you can save someone, or you can kill a lot of people for no reason. It's not going away, that's for sure. So we'll just have to learn how to live with it.