monty

Bitcoin. NFTs. Artificial Intelligence. It feels like every couple of years, a newfangled technology leaps to prominence following the sudden enrichment of a few lucky people. Unfortunately, those lucky people tend to be unlikeable libertarians. This leads to a tendency among people who enjoy arguing with unlikeable libertarians to dismiss the technology outright. Admittedly, I am guilty of this. There may be or there may not be useful applications of the blockchain that aren't pyramid schemes built around collecting pixel art of chimpanzees, but the objectionable aesthetic of the people advancing the technology makes me biased against the core technology itself.

AI is different. The utility of AI technologies is obvious to anyone who isn't sticking their fingers in their ears or burying their head in the sand out of fear. AI tools like ChatGPT are being used and abused by millions of people. Image generation technology is producing images of increasingly astounding clarity. Every few months, a new tool exponentially more capable than what was being used even a year before is announced. Unlike NFTs – which have died, and Bitcoin – which lives in an endless yoyo of collapse and recovery, AI has already begun to be integrated into the social fabric in a way that makes it obvious that it is not going anywhere. Yet despite this widespread adoption, AI seems to fill many people with a not totally unreasonable dread. Digital artists in particular seem to be the most vocal critics of this technology. Their reactions to it vacillate between desperate fear of replacement and smug disdain towards its supposed incapability. In this article I'm going to address and respond to the response to AI from professional artists and their friends and allies: what they are right about and what they get wrong.

Are AI images art?

The most common criticism levied by artists is that AI is incapable of producing art. I suppose this opinion is subjective, as art is difficult to define, but I would strongly agree. AI is not an “artist” or a producer of art in itself, but a tool of production.

One of my favourite art commentators, Brad Troemel, points out that art fundamentally is an expression of subjective experience. In Troemel's opinion, art flows from the artist's relationship with and perception of the world around them. Each artist's style is unique and usually distinguishable, even if only subtly. I think this is a fairly intuitive statement. It's why diversity is valued so highly in art in a way that it isn't in warehouse work or manufacturing line supervision. The art world generally wants there to be a large range of subjective experiences among artists as it leads correspondingly to a wide range of artistic expression. The fact that AI “art” generators tend to produce images, poems, and stories that mimic human art implies that the AI is not engaging in art as an artist. While it is technically possible that the subjective experience of a diffusion model or a large language model is similar to a human's, I seriously doubt it. If AI were engaging in art as a producer rather than as a tool of production, we should expect it to produce something different from human art, possibly something incomprehensible or more akin to the average output of Deep Dream than Stable Diffusion. As it exists now, generative AI merely mimics what a probability model expects a human to produce. If a human is not involved in the design process of a piece, an AI art piece is merely a simulacrum of human expression.

To be art, a piece must be made with AI instead of by it. The qualitative distinction here is frustratingly unobjective. If I were to draw a stick figure scene and feed it into an image-to-image model, would the product be made by or with AI? If I draw 99% of a complete landscape and then use AI to generate a tree in the background of the image, is the final piece made by or with AI? In the latter case, how is the use of AI fundamentally different than the use of another assistive tool like a ruler or a compass? YOU may have hard answers to all of these questions, but that doesn't actually matter. The distinction between “too much” and the “right amount” of AI assistance in the artistic process will vary between different producers and consumers of art.

Are artists replaceable?

Artists aren't wrong to make the point that pure AI images are not art and I support that position, it's just irrelevant to the debate around generative AI. This point exists as a mask over the true dispute between digital artists and AI, which is that AI is a nearly apocalyptic threat to artists as a social class. In accusing images made with or by AI of being “fake” art, they are not motivated by some instinct to snobbishly gate-keep but by a material need to convince consumers of art that what they produce can not be replaced.

As a commodity, art has 2 broad use values. The first is art for its own sake. There is an “art world” of artists, galleries, and collectors that produce and consume art as a commodity due to the perception of art production as a noble pursuit. To the people in this world, the idea that AI could replace artists is farcical. Art exists to expose observers to a unique subjectivity. Fundamentally, the logic of the art world is an unusual condition of production. Artists in this sphere are monopoly producers, the value of their art determined largely by the perception of them as a brand. A piece by Banksy is more valuable than your average street artist not because of the socially necessary labour time of Banksy's work but because art by Banksy is a class of commodity in and of itself.

The art world only makes up a tiny fraction of art consumption. Most consumption of art falls under art's other use value: representation of a concept in the world at large. Images drawn by humans exist as book covers, advertisements, warning signs, representations of fictional characters, and even, unfortunately, pornography. The consumer of this class of art commodity is everyone. It is impossible to live in the 21st century and not consume some subset of this second class of art as a commodity. The vast majority of expenditure on this specific class of commodity art is done not for the love of art in and of itself but because the art serves some sort of social purpose.

It is true that AI can never replace artists whose art lies largely within the first category, but most working artists fall into the second category. They live on commissions or contracts or as designers within a corporate structure. It is this second category of artists whose position is threatened by AI image generation. Because their art is only art incidentally and is instead consumed for a purpose outside pure appreciation, their work could conceivably be replaced with an image that is not art, as long as it is a close enough approximation to still sufficiently fulfill the social purpose a human artist would've been hired to meet. It is certainly true that there are people who appreciate the artistry of book covers, print advertisements, and even road signs, but this appreciation is not why those forms of visual art are manufactured.

Of course, this is only a threat if AI is capable of fulfilling the social need common art fulfills. Digital artists' second refrain after “AI art is not art” is usually “Even if AI art is art, it isn't good art.” This isn't unfounded. There still exists a gap between human artists and AI art, but complaints made today are often “this high-quality image of a man has an extra finger” and not the complaint that “this image is not recognizable as a man” that people were making only a year or two ago. The progress of AI image generation has been extremely rapid and there isn't a good reason to believe we've reached a plateau in capability. Take for example the popular generator MidJourney. These are pairs of images generated from the same prompts one year apart:

“A superhero like Batman with a dark red futuristic cardinal themed costume. He has a mask with a beak. In comic book style.”

“A female teenage superhero. She has red hair. Her costume is Italy-themed. She has an Italian flag for a cape. In comic book style.”

AI image generation has progressed so incredibly fast that as early as 2022 it was found that image generators could literally be trained to use human brainwaves as a prompt to reconstruct an imagined image. (Takagi, Yu and Nishimoto, Shinji, 2022)

Image generation threatens the livelihood of the common artist and threatens to destroy them as a social class. This is the real source of animosity between artists and AI. The existence of AI image generation lowers the labour value of art as a commodity for its use value as a representation of a concept to well below the cost of hiring an artist to perform the same social function. The answer to “are artists replaceable” is in many cases, yes.

The future of artists

If you asked the average person whether or not they would rather buy a book that has a cover drawn by a human or an AI, they would probably say a human. But what about when it becomes impossible to tell the difference? What about if a book with an AI-drawn cover is a dollar cheaper? If you asked the average person if they would rather buy food grown by a local farmer or a farmer on the other side of the planet, they will probably SAY local but buy cheaper foreign produce anyway. When actually confronted with the consumer decision the cheaper option is chosen more often as long as it meets the same need.

To an outsider, the fight between artists and AI seems less like an ideological dispute between “pro” and “anti” AI and more like the impotent riots of the Luddites in Britain between 1811 and 1817. Just as the Luddites were unable to stop capitalism's tendency to concentrate wealth, proletarianize, and increase productivity so too will artists be unable to roll the world back to how it was before the AI explosion. As @h has said about this very topic, “the genie cannot be put back in the bottle.”

Even though artists are staring down an AI-generated image of a barrel of a gun, art will continue to be made. Artistic expression is a part of the human condition. Art will still be produced as a hobby and there will still be (a much smaller group of) artists able to survive by selling their art to common people, just as there are still people who sell artisanal candles and knit hats even though 99% of candles and hats are now mass produced.

There will also continue to be Banksys and Gerhard Richters. Artists whose art is a luxury product consumed within the community of the “art world.” These people will probably come to embrace AI as a useful tool like their predecessors did with the once unprestigious acrylic paint and plastic paint brushes. However, with the prospect of a professional career in art becoming even bleaker in a post-AI world, more of these people will be sons and daughters of the wealthy and connected within that already insular community of fine artists.

Capital concentration is inevitable and immiseration comes for us all. Artists are no exception.

Coming soon

It's been a long time since I published something on the printhouse. That's because I've fallen into a bad habit of starting research on an article, planning the article, and then coming up with an idea for a different article and starting all over again. As a result, I have an ever-clogging pipeline of articles in production. Look for the following coming in the indeterminate future: – Range Feudalism 2: If they are victims of this tendency, why do farmers tend to support this? – Software Exchange Value redux: my original article contains a fundamental error so I unpublished it. I'm reworking it and when it is released again, it's going to be very different from how it was before. – Automation: a follow-up to this article about the (im)possibility of AI technology totally replacing the proletariat. – Culture War: what is it and where does it come from? Is there any escape? – Deep learning From First Principles: an explanation of how exactly AI works assuming nothing but high school math knowledge.

I think in a previous article I promised to write Rage for the Machine, a history of the CIA's infiltration of anti-establishment/counter culture, but I've scrapped this article because a lot of the evidence is simply too schizophrenic or dubious.