In 1903 James Joyce defined art as "the human disposition of [...] matter for an aesthetic end", i.e. created by someone consciously manipulating materials solely to create beauty. It's a straightforward, seemingly bulletproof definition—except Joyce then defines photography (a non-human disposition of matter) and furniture (which exists for practical, not only aesthetic, ends) as not being art.
Most efforts to definitionally exclude AI from being considered "art" get tied into the same knots, a problem I feel your approach of emphasising intentionality and process helps sidestep, focusing not on a narrow technical definition but on the experience of creating and engaging with beauty. I don't agree with every point (I can accept some forms of artifice/advertising as falling under the "art" umbrella), but it echoes my own objection to AI "art"—that the pleasure of art exists in the challenge I have to do in expressing ideas beautifully or making sense of them in someone else's creation, and taking the difficulty out of it takes away the pleasure.
It's interesting to see the way AI, specially image and video generation tools, has evolved in the last 4 to 5 years. The most distinctive evolution, I think, is the transformation of AI from a tool of experimentation to a tool of mass production. If one looks to early "AI artifacts" (such as the works from Helly Herndon or previous works produced by image generation tools such as ArtBreeder or Generative Adversarial Networks (GAN)) one can see that the difference between input and output was HUGE. The machine was truly stochastic, random, esotheric and confusing. Those dimensions — of what I would describe as "machine unconcious" — were the most prominent dimensions of any AI work, making them truly baffling and eary. In other words, they seem to reflect something uncontrollable in those models. In those cases, the artist was a truly curator, I think.
Nowadays AI in general is commercialized as a tool for production. For a process that NEEDS to be standarized and not stochastic. In any case — be they words or images — the result must reflect an intention, and must resemble what ones asks for.
To think about these early examples of AI helps us, I think, to broaden our understanding of the capabilities of these tools as creative tools, and also help us articulate the dimensions that are lacking in current AI models.
These reflections also remind of a beautiful piece from L.M Sacasas that talks about AI being "lonely surfaces" in contrast to surfaces (images) being able to connect us with someone, a surface "that arouses the desire to know them and the depth to fulfill that desire.” Here's the piece: https://theconvivialsociety.substack.com/p/lonely-surfaces-on-ai-generated-images
I was anxiously awaiting your argument getting to the case of photography, and was surprised you didn't mentioned it explicitly.
Whenever I catch myself frowning at AI slop, the way you describe at the beginning of the article, I inevitably draw the analogy to the advent of photography, and Baudelaire's reaction to it (as well as many others').
Not sure photography really passes the threshold of art according to Matel's distinction. So is all photography artifice? Or do we have to water down the distinction, which inevitably opens the door for AI Slop?
You are correct that photography is probably the best recent case study, and I thought about including it as an example, but it's quite complicated I think and the essay was already getting pretty long. But I'll attempt to sketch out some of my thoughts about it here.
First, can photography be art? I think photography tends towards artifice but can certainly be art. If you simply compared someone casually snapping a landscape photo, to someone painting the same scene, the photograph certainly falls much closer to artifice most of the time. However there are still many choices that the photographer can make in the process of capturing a photograph that I think allow for the artistic process to sufficiently manifest. (You can chose a model, pose them, choose a location, time of day, direct the performance, light the scene, choose the composition, shutter speed) these things together present plenty of opportunity for the artistry of photography to present itself through the work even though it may be less laborious than producing a similar image via painting. However, photography is certainly a technology that enables the creation of much artifice, in a way that painting didn't.
So what then is the difference between photography and AI images? I'd argue that "creating" images through AI even further reduces the number of choices the human makes, and amount of control the human has, providing even less space for artistic expression to manifest itself through the work. You could argue that you can control things like pose, location, framing, etc through the prompt, but I think there's a meaningful distinction to be made between those things being the result of a process of creation (as with photography), and merely a process of description, where something else does the creation for you to achieve that result (as with AI gen). The line between these two is slimmer, and less obvious, but I think still great enough to make a distinction. And if photography is a technology where art is still possible but the technology pushes us closer to artifice by default, then AI gen is even further along that spectrum.
Your ability to formulate an argument, eloquently deconstruct it, and provide constructive feedback as to possible alternatives without taking a hard stance is so inspiring. A film analyzer/ teacher/ philosopher.
In 1903 James Joyce defined art as "the human disposition of [...] matter for an aesthetic end", i.e. created by someone consciously manipulating materials solely to create beauty. It's a straightforward, seemingly bulletproof definition—except Joyce then defines photography (a non-human disposition of matter) and furniture (which exists for practical, not only aesthetic, ends) as not being art.
Most efforts to definitionally exclude AI from being considered "art" get tied into the same knots, a problem I feel your approach of emphasising intentionality and process helps sidestep, focusing not on a narrow technical definition but on the experience of creating and engaging with beauty. I don't agree with every point (I can accept some forms of artifice/advertising as falling under the "art" umbrella), but it echoes my own objection to AI "art"—that the pleasure of art exists in the challenge I have to do in expressing ideas beautifully or making sense of them in someone else's creation, and taking the difficulty out of it takes away the pleasure.
It's interesting to see the way AI, specially image and video generation tools, has evolved in the last 4 to 5 years. The most distinctive evolution, I think, is the transformation of AI from a tool of experimentation to a tool of mass production. If one looks to early "AI artifacts" (such as the works from Helly Herndon or previous works produced by image generation tools such as ArtBreeder or Generative Adversarial Networks (GAN)) one can see that the difference between input and output was HUGE. The machine was truly stochastic, random, esotheric and confusing. Those dimensions — of what I would describe as "machine unconcious" — were the most prominent dimensions of any AI work, making them truly baffling and eary. In other words, they seem to reflect something uncontrollable in those models. In those cases, the artist was a truly curator, I think.
Nowadays AI in general is commercialized as a tool for production. For a process that NEEDS to be standarized and not stochastic. In any case — be they words or images — the result must reflect an intention, and must resemble what ones asks for.
To think about these early examples of AI helps us, I think, to broaden our understanding of the capabilities of these tools as creative tools, and also help us articulate the dimensions that are lacking in current AI models.
These reflections also remind of a beautiful piece from L.M Sacasas that talks about AI being "lonely surfaces" in contrast to surfaces (images) being able to connect us with someone, a surface "that arouses the desire to know them and the depth to fulfill that desire.” Here's the piece: https://theconvivialsociety.substack.com/p/lonely-surfaces-on-ai-generated-images
I was anxiously awaiting your argument getting to the case of photography, and was surprised you didn't mentioned it explicitly.
Whenever I catch myself frowning at AI slop, the way you describe at the beginning of the article, I inevitably draw the analogy to the advent of photography, and Baudelaire's reaction to it (as well as many others').
Not sure photography really passes the threshold of art according to Matel's distinction. So is all photography artifice? Or do we have to water down the distinction, which inevitably opens the door for AI Slop?
You are correct that photography is probably the best recent case study, and I thought about including it as an example, but it's quite complicated I think and the essay was already getting pretty long. But I'll attempt to sketch out some of my thoughts about it here.
First, can photography be art? I think photography tends towards artifice but can certainly be art. If you simply compared someone casually snapping a landscape photo, to someone painting the same scene, the photograph certainly falls much closer to artifice most of the time. However there are still many choices that the photographer can make in the process of capturing a photograph that I think allow for the artistic process to sufficiently manifest. (You can chose a model, pose them, choose a location, time of day, direct the performance, light the scene, choose the composition, shutter speed) these things together present plenty of opportunity for the artistry of photography to present itself through the work even though it may be less laborious than producing a similar image via painting. However, photography is certainly a technology that enables the creation of much artifice, in a way that painting didn't.
So what then is the difference between photography and AI images? I'd argue that "creating" images through AI even further reduces the number of choices the human makes, and amount of control the human has, providing even less space for artistic expression to manifest itself through the work. You could argue that you can control things like pose, location, framing, etc through the prompt, but I think there's a meaningful distinction to be made between those things being the result of a process of creation (as with photography), and merely a process of description, where something else does the creation for you to achieve that result (as with AI gen). The line between these two is slimmer, and less obvious, but I think still great enough to make a distinction. And if photography is a technology where art is still possible but the technology pushes us closer to artifice by default, then AI gen is even further along that spectrum.
Your ability to formulate an argument, eloquently deconstruct it, and provide constructive feedback as to possible alternatives without taking a hard stance is so inspiring. A film analyzer/ teacher/ philosopher.