Yeah, it’s not so much about the difficulty or the practice, it’s in the innate lack of precision that AI offers to manifest any meaningfully specific artistic vision. Whatever vision the creator might have has to be compromised to accept the closest match AI could come up with.
A definition of art that I liked basically said that art is about creative expression through a medium. Whether that expression is done poorly or well is irrelevant. AI “art” doesn’t really express anything except “here’s a possible thing I could make in response to this prompt”.
The prompts themselves are more creative and more artistic than any of the outputs, and I wonder if they could even be copyrightable.
Right, sounds like someone writing a book versus making a movie adaptation. The book has me fill in the gaps, which can in many cases be more satisfying than whatever the film adaptation comes up with, but sometimes a film adaptation executed well adds something. However a hypothetical book-to-film AI would be utterly mind numbingly uninspiring, and I would just take the prompt and use my own imagination to generate what I’d like more.
It’s been fairly obvious to me for some time that a website where people shared the prompts instead of (or alongside of) the output, and that allowed you to hit a generate button to get different versions would be way fucking cooler than a seemingly infinite scroll of uninspired, uncanny, single generation outputs.
I’m conflicted here. Your first paragraph initially opens the loophole “but AI art” could be the medium in which the artist is expressing themselves. So poor prompting could be beautiful too, in a way. I’m sure photographers of the past felt this way about software post editing when that became popular.
The results may be good to many viewers but apalling to anyone who can tell the difference. If the results don’t matter, does it matter if AI slop is “bad”?
The prompt may have been beautiful, and the process of learning and finding the right tools (i.e. choosing the right model) is akin to the struggles of any artist in learning their craft.
If the results don’t matter, does it matter if AI slop is “bad”?
I mean I didn’t say all art defined that way would be good art. But I also don’t find the outputs artistic at all.
The prompt may have been beautiful, and the process of learning and finding the right tools (i.e. choosing the right model) is akin to the struggles of any artist in learning their craft.
I was afraid going down this road too far would make me sound like I was an advocate for this technology. So I typed out something similar multiple times in my initial reply and then deleted it, but that was kind of my thinking as well.
I read an article where a writer was using autocomplete – in the before times when it wasn’t called AI – to try to write a piece about a relative that recently died that fully included the usage of the tool in the article itself, and that’s the closest thing to AI assisted art I’ve ever read. Now I wish I had saved it.
With a human, a limitation in skill or material properties will tend to manifest in a vague way that lets me fill in the gaps, or else leave non-essential details out.
With GenAI, some details will generally be added, but without any intent behind them.
As another commenter pointed out, I’d probably rather read the prompt and treat it like a book rather than look at the GenAI output stemming from that prompt. If there’s no voice acting, With text content I can just read it instead of looking for meaning in the voice acting when there’s no intent behind the voice. With visual art, then I know there’s intent behind whatever details are there and don’t need to pay too much attention to stuff that doesn’t matter.
Yeah, it’s not so much about the difficulty or the practice, it’s in the innate lack of precision that AI offers to manifest any meaningfully specific artistic vision. Whatever vision the creator might have has to be compromised to accept the closest match AI could come up with.
A definition of art that I liked basically said that art is about creative expression through a medium. Whether that expression is done poorly or well is irrelevant. AI “art” doesn’t really express anything except “here’s a possible thing I could make in response to this prompt”.
The prompts themselves are more creative and more artistic than any of the outputs, and I wonder if they could even be copyrightable.
Right, sounds like someone writing a book versus making a movie adaptation. The book has me fill in the gaps, which can in many cases be more satisfying than whatever the film adaptation comes up with, but sometimes a film adaptation executed well adds something. However a hypothetical book-to-film AI would be utterly mind numbingly uninspiring, and I would just take the prompt and use my own imagination to generate what I’d like more.
It’s been fairly obvious to me for some time that a website where people shared the prompts instead of (or alongside of) the output, and that allowed you to hit a generate button to get different versions would be way fucking cooler than a seemingly infinite scroll of uninspired, uncanny, single generation outputs.
I’m conflicted here. Your first paragraph initially opens the loophole “but AI art” could be the medium in which the artist is expressing themselves. So poor prompting could be beautiful too, in a way. I’m sure photographers of the past felt this way about software post editing when that became popular.
The results may be good to many viewers but apalling to anyone who can tell the difference. If the results don’t matter, does it matter if AI slop is “bad”?
The prompt may have been beautiful, and the process of learning and finding the right tools (i.e. choosing the right model) is akin to the struggles of any artist in learning their craft.
/devil’s advocate
I mean I didn’t say all art defined that way would be good art. But I also don’t find the outputs artistic at all.
I was afraid going down this road too far would make me sound like I was an advocate for this technology. So I typed out something similar multiple times in my initial reply and then deleted it, but that was kind of my thinking as well.
I read an article where a writer was using autocomplete – in the before times when it wasn’t called AI – to try to write a piece about a relative that recently died that fully included the usage of the tool in the article itself, and that’s the closest thing to AI assisted art I’ve ever read. Now I wish I had saved it.
You could say the same about limitations in skill or material properties or…
With a human, a limitation in skill or material properties will tend to manifest in a vague way that lets me fill in the gaps, or else leave non-essential details out.
With GenAI, some details will generally be added, but without any intent behind them.
As another commenter pointed out, I’d probably rather read the prompt and treat it like a book rather than look at the GenAI output stemming from that prompt. If there’s no voice acting, With text content I can just read it instead of looking for meaning in the voice acting when there’s no intent behind the voice. With visual art, then I know there’s intent behind whatever details are there and don’t need to pay too much attention to stuff that doesn’t matter.