Say it A.I.n’t So!!
A round-up of creative industry tea surrounding AI
Well howdy!
Hope everything has been going stupendously for you over the last two weeks. Over here I’ve been tying up loose ends on some projects and exploring old haunts during a brief trip to Dallas.
This relaxed change of pace afforded me some time to visit with friends, do a little outreach, and catch up on creative industry tea—I mean news. Which brings me to our catalysts for this week’s slightly unconventional newsletter: AI-related drama in the UK Parliament and a recent article about contracts, copyright, and AI image harvesting published by It’s Nice That.
Can artists get a little R.E.S.P.E.C.T. please?!
Let me preface this (before the tech bro stans come for me) by saying that, for the most part, AI as a tool is intriguing. I appreciate the capacity it has to speed-up workflows, aid in scientific research, improve graphics and summarize information—among other functions. It has great potential to assist humanity.
But, on the other hand… we live in a world riddled with powerful, opportunistic actors who would happily replace human labor for the sake of a buck (or like, billions of bucks). For example: if a media conglomerate wants to own your IP—permitting them to use it for anything—they can train AI to iterate that art into infinity and never hire you again. And when that’s offered in exchange for a single, small editorial commission… it’s hard to call that anything but greed in its most absurd form.
But this is not the action of some cartoon villain. This is precisely the predicament that illustrator Aysha Tengiz thwarted earlier this year. The deeply predatory nature of this situation prompted her to share the experience on Instagram along with her newfound (and necessary) obsession with copyright law. As she writes in her subsequent article for It’s Nice That:
The parent company of the magazine that originally reached out to me is massive. If this is a standard contract that gets sent to every artist that works for them, then they may have an enormous hub of artwork that they can train AI with. This means they could use AI to repeat the artwork style by referencing these images, without having to re-hire the artists who created the work. And it would also be completely legal, because they technically own that art.
When Tengiz countered with an offer of “an indefinite license to use the work in the magazine”—which still allows for incredible flexibility in usage rights—the company maintained their initial offer. So, she turned down the project. Good for her! Though, in an era where more creative work (and money) is taken from the hands of artists and fed into computers, that decision isn’t made lightly.
Creative Sentiments
I’ve been met with a mix of responses when it comes to creatives on AI. Some people find it exciting, using it as a tool for quickly ideating and iterating. Others brush off the threats to our industry with a (somewhat naive) “AI sucks, it will never be able to replace a human.” Or “Well, if a client isn’t going to value your humanity and vision anyway, do you really want to work with them?” And these are all fair responses. But… I feel that they miss the point.
The point is that there are few regulations governing these increasingly sophisticated systems, which means artists have limited protections. Tech companies have already scraped the Internet, taken our work to train their image generators, and are directly competing with us. Even Jessica Hische, one of the leaders in our field, admits in a LinkedIn post that she is not immune to the negative impacts of AI.
Legal Happenings
As we await definitive legal rulings in the US to determine whether or not taking copyrighted materials to train AI is fair use, we at least got word that (for now) art created by AI is not copyrightable. And a recent US Copyright Office pre-publication opinion hints that tech giants might not be able to cry “fair use” for everything. An excerpt (from p.111):
But making commercial use of vast troves of copyrighted works to produce expressive content that competes with them in existing markets, especially where this is accomplished through illegal access, goes beyond established fair use boundaries.
Of course, the head of our Copyright Office was fired two days later. Go figure.
Over in the UK there has been a back and forth between the House of Commons (evidently aka “the government”) and the House of Lords regarding an amendment to the Data (Use and Access) Bill. As the BBC explains, this addition would “add transparency requirements, which aim to ensure copyright holders have to give permission for their work to be used.” The amendment has received overwhelming support from British creatives, which is why—as the House of Commons continues rejecting it—celebrities like Elton John are compelled to call their asses “losers.”
Anyway, I feel the need to abruptly stop myself from going any further down this rabbit hole. But the impact—positive and negative—of AI on our field is something artists must stay informed about, which is why I bring these news stories to your attention. As Aysha Tengiz remarked in her article, “Clearly, it’s now more important than ever for artists to understand and protect our rights.”
Helpful links
The US’s Graphic Artists Guild provides some free resources regarding copyright for artists, as well as the UK’s Association of Illustrators.
If you would like to “poison” your art as an offense against AI image generators, you can try out Nightshade. Another option is Glaze, though this defends your art style without the poisoning part (NPR explains a bit about how they work).
This article actually got me curious about playing with text-prompt image generators—particularly the section about art director Karen X Cheng’s Cosmo AI cover art process.
Yury Vetrov explores the many use cases of AI across the design field. I particularly appreciated the “What To Read” section.
Spawning.ai has created some defensive tools for artists to use on their websites, including Kudurru (blocks AI scrapers) and ai.txt (tells machines not to train on your content).
You can see if your art is in the most popular AI image dataset, LAION-5B. This browser extension also helps identify images that have been scraped to train AI. Both are offered by Spawning.ai.
The Verge covers some of the above recommendations, as well as tips on opting your work out of training datasets.
Note: Some of these resources I’m still in the process of exploring myself, so if you’ve used any of them please let me know your thoughts!
Well, that’s it for this week’s edition. Unless some bombshell drops in the next two weeks, I’ll be back to sharing studio updates and art in my next newsletter.
As always, feel free to reach out with questions and comments. Since this is a somewhat hot button topic I only ask that you be nice (please)!!
Ciao ciao,
-Liz
This is so well done Liz! Really appreciate your thoughtful, balanced, thorough exploration + so many great resources! (P.S Sir Elton for the win!)