Ok, I’m going to write one last thing about AI Art. This time, with no trolling or sassy memes. So I hope y’all can read it with an open mind.
So there was considerable pushback to my previous post and there were some very valid POVs. To the people who said stuff like “AI is being used unethically and I don’t want to f_ck with it,” Yes! You are right to feel this way and I respect that position. It’s equally valid to say “AI is being used unethically, so I want to explore the limits of this technology and how it affects the world we live in.” and it’s also equally valid to say “AI allows me to concept and source visuals that I can then process into my own artworks.”
But here’s the thing. There is more than one kind of visual art practice and many of them are not existentially threatened by generative image AIs. If anything, the artists who are threatened by this tech are the ones who make a living off of making representational imagery for a number of cultural industries. This in itself is a privileged position that isn’t afforded to most visual artists. Many artists who exhibit and publish around the world don’t get the opportunity to live off of their work, they have to teach or work in adjacent industries to survive. This is especially true for new media artists like myself.
That said. It’s obvious that these generative image AIs are quite adept at simulating visual data. But anyone who has used Midjourney or Dalle2 can attest, there’s so much that these AIs can’t and won’t do. Visual art is not just the mechanical simulation of visual style.
Here’s what generative image AIs can’t do, it can’t create images with a sense of:
Concept
Materiality (Physical or Digital)
Criticality
Yes, it also can’t create hands and it’s not that great at drawing Black people either. But more importantly, AI can’t imagine what it means to work with materials, whether it’s the scratched up photos of Adrian Piper, or the glitch artifacts of a specific image format. Generative image AIs can’t create conceptual art and it can’t look at or arrange images in a critical way. What I’ve noticed is that many artists (not all ofc) who work on these levels of meaning are not threatened by AI, many of them see it as an interesting tool that they may or may not use. Working with these layers of meaning is what keeps artists ahead of the curve in terms of how these automated technologies are developed.
There were similar hysterics with the advent of photography, and it took many decades for photography to be accepted as an art form. “You just have to push a button,” they’d say. “Photography is going to kill the fine art of painting.” yada yada. Photography certainly didn’t kill painting, if anything, painting was freed from the burden of representation. We wouldn’t have cubism, or suprematism, or abstract expressionism, or pop art without the development of photography. It’s true that many portrait artists lost their gigs with photography, but overall humanity’s creative drive survived and flourished.
(Side note: Generative image AIs are going to eff up photography in many ways, some scary, and I think as a post-photography artist, it’s important to interrogate these dynamics rather than disavow it. But that’s my imperative. Stock photogs and some fashion photogs are going to get screwed, but not photojournalists or documentary photogs as much.)
So yes. Generative image AIs are basically appropriation machines and, as far as appropriation machines go, its like the equivalent of a visual art sausage maker. It averages all the images scraped from the internet into a kind of fast-food aesthetics.
But here’s the rub. Generative image AIs can’t imagine what doesn’t already exist. Every output is the result of averages from existing visual data . AI will never replace an unfettered imagination. This is another reason why the whole NO AI ART movement doesn’t speak for all artists, because many of us are striving create images that haven’t been created before.
But I get it, this is Tumblr, and since the great purge of 2018 the dominant aesthetic has been fandom-based art, which relies on borrowed aesthetics much in the same way that AI does. I think the reality is that many artists who work in this niche are going to get royally f_cked and my heart goes out to them. It won’t be the AIs that do the f_cking, but the IP owners like DisneyMarvel, Sony, Epic Games, Nintendo, etc . They will be the first to use generative image AIs so that they can save money by not hiring artists. This will be especially tragic for the privileged few who went to art school to land industry jobs.
These existential threats to artists are ultimately a question of labor rights as opposed to technology.(UNIONIZE!) Appropriation is literally the engine that drives all art from the beginning of time. Every artistic tradition has grown because of appropriation. Whether it’s blues or classical music, Shakespearean tragedy, renaissance paintings, pop art, conceptual art, hip-hop, you name it, there has been appropriation. The question about whether appropriation is good or not, is a question of where capital is flowing to. Like Rock n Roll is crap because it channeled capital from a Black musical tradition to the hands of white musicians. But hip-hop was great because sampling (a real art form, like AI art) reversed that trend in that you could repurpose white people music into a new urban Black musical tradition. Ofc this is a simplistic summary, look at how long it took Biz Markie to get royalties, etc
So yes, Generative Image AI Services are appropriating art works in ways that are harmful to some artists and these artists should get royalties every time their name is conjured up in a text prompt. But I’m sorry, this doesn’t mean that “artworks are being stolen.” Nothing has been stolen.
What people don’t realize was that many of these image models were created years ago. Data was scraped from the internet on a mass scale and this was all legally sanctioned because of Fair Use Laws that protect researchers who want to scrape data. These datasets, like LAION-5B which itself is based off the Common Crawler dataset, are open source and freely available. They were created for research purposes within research settings and that is totally legitimate. Fair Use Laws are essential to artists and data scraping is one of those things that can be used for good or evil. But the way these AI models were trained was completely legal. What is ethically dubious and f_cked up is how corporations used these legal mechanisms designed to help artists and researchers to create a privatized service product. Again, this is why I say, “you don’t hate AI Art, you hate capitalism”
The reality is that an artist can copyright an image, but not a general style. This idea that every artist on ArtStation has some inalienable originality and uniqueness to their work is the result of a hyper-individualistic capitalist ideology that is utterly out of touch with how art has been historically produced. Since the advent of mechanical reproduction, the magical aura of the art work has long been dead. Authenticity and originality are marketing buzzwords and not facts.
These “AI stole my art” arguments are crouched in a deeply regressive view of copyright and intellectual property, because very few artists, if any, would be able to prove in court that an AI plagiarized their works. AIs are sophisticated averaging machines, so the more an artist’s work appears to be copied, the more datapoints of their works and similar works exist in the training model. Proving that AI “steals art works” would also mean dismantling Fair Use Laws, which actually protect artists from predatory corporate IPs.
Digital remix culture, fan art, hentai, all the things you dearly love and breathe on this Tumblr site would evaporate without Fair Use and it’s hella naive to think that if one of these illustrators actually proves that AI “stole their work,” that Disney and Studio Ghibli and Nintendo won’t come around and say that illustrators are in turn stealing their precious IPs. Copyright laws exist to protect the ruling class. There’s a reason why all us lefty new media artists release works as CC. (Side note: Still waiting to see an artist from the Global South make a “AI steals my arts” argument, so far i’ve only seen US and EU-based artists fret.)
And there is a sad irony in all of this. None of my works were used to train AI even though Tumblr was among the sites that were scraped. (I never used ArtStation or Deviant Art bc they don’t eff with experimental media artists) The reason is that there’s a preference for images that are algorithmically legible. The effed up reality of internet platforms is that algorithms judge your images before distributing them. Glitch art and experimental media arts are usually judged to be low quality content by the instagrams and tumblrs of the world. I had like 200 abstract images flagged during the great purge of 2018 that I appealed and won, because these algorithms don’t have a clue about what is art and what isn’t. Everything on ArtStation, stock image websites, are easily legible by nature, and as such, easy to algorithmically appropriate. Artists who know how to create works that are algorithmically illegible will thrive in the post-AI era.
And this leads me to my final point. Many artists, like myself, are drawn to AI for what I call “Liminal Aesthetics.” It’s not about how good the AI can approximate an image, but how bad it is. Many of us are interested in these mal-nourished images. The weird faces and hands, the lack of proportions, the weird shapes, and unexpected textures. The thing though, is that as these generative image AIs get more advanced, the liminal aesthetics are lost. The more these AIs get better at parsing natural language, the more basic and unaesthetic the outputs. So for many artists, this AI moment is temporary. As DALLE2 and MidJourney will ultimately turn into a fancy clip art generator that will replace stock image and stock graphics sites, from iStock to Sketchfab. That’s why I think it’s important to embrace this shit now, because these gorgeous liminalities will soon be a historical artifact.
To conclude, I will say this, and this is very important. The trad art world and their system of values is dying. In the post-digital world of infinite images that aren’t bound by physical scarcity, the rarest and most valuable images are the ones that are remembered, shared, and viralized. If your art is being copied or replicated, even by an AI or a bot, that means that your art is surviving this new era. There are worse fates than having your art copied.
That worst fate is censorship and forgetting.
The image at the top is [ILLUSTRATION #4j2492D] by jonCates, he describes ths AI work as “The Artificially Illustrated Glitch Western Primer for Machine Learnerrs,” and it’s part of a much broader project of creating the first “Glitch Western” based on historical narratives of Black and Chinese people in the West. Most of this project isn’t AI-based imagery, but lovingly crafted video and game art.
Some may not know this, but jonCates was big on Tumblr before the purge of 2018. He’s an og glitch artist and his “Dirty New Media” tumblr was a veritable museum of lewd glitch arts. I remember being so proud when my boo’s glitched bootie ended up on that page. The site was a treasure. Sadly though, Tumblr deleted all record of the Dirty New Media blog as well as jonCates’ tumblr of personal work. All f_cking gone forever. This is the real tragedy of censorship and forgetting. A whole culture lost.
So why cling to this destructive dynamic because you read a hot take about AI by a privileged artist working in gatekeeping industries? What is copied and appropriated can still live on and we all can do better and fight the power where it resides instead of vilifying other artists. Cheers, shout out to everyone who read all this <3
A nuanced (but direct and clear-eyed) read that is utterly worth your time.
