currently, the biggest thing visual artists can do to make obvious the difference between intentional art and diffusion slop is to create clean lines and distribute at a high resolution.

Follow

@akjcv idk, i think the biggest thing is just to create art which is intentional

@Lady I do think it's interesting to interrogate which specific visual techniques are made difficult, not by particular technical shortcomings in the models that exist today, but by the technique of diffusion as a whole

Like I get what you're saying, and I don't disagree, but I do also think that there are things that result in obvious visual distinctions and it's not cheapening the art or whatever to talk about that

@akjcv sure, it is interesting from an art history perspective, but i’m kind of biased in terms of like, what we do with that

like, “due to the presence of A·I generated visual media and its perceived tackiness, artistic styles and techniques which reminded viewers of A·I fell out of favour, while artistic styles which did not resemble A·I output became trendy” is an interesting statement. there’s a lot to unpack there.

but i also think it’s bad. it IS a reduction in our artistic space and a cheapening of art to value it based on whether it “looks like” a certain kind of generated visual media instead of judging it on its own merits. and in particular: clean lines and high resolutions are labour-intensive and difficult to produce. it’s hurting a lot of artists if we assign a lot of value to that.

i’m not saying you ARE assigning value in this way, but it’s the inevitable endpoint in a society which DOES assign value to artwork based on these factors. and i wonder if it is necessary. there are other things generated art is bad at: semantic coherence; intentional artistic decisions; communication of sentiment. why are we indexing on the technical aspects of the artwork and not these human- and art-oriented things? isn’t that kind of playing the game by A·I’s rules instead of our own?

@akjcv the entire webcomic era was low-resolution images with messy lines. and yet i have never seen an A·I produce a coherent webcomic. so your original claim warrants further investigation.

@Lady sure. and isn't it interesting to think about why it can't? I think it's partly because of the necessity of repeating recognizable characters, partly because of the extreme demands the format makes in terms of consistency and space partitioning, and partly because of how conceptually complex webcomics tend to be.

i don't think it's enough to just say, "make art with intention" and leave it there. there are specific things these tools can't do and it's useful to know what they are.

@akjcv i don't think you DO have to say “make art with intention”. i think what you need to say is “learn how to recognize intention within art”. A·I art is only compelling because art criticism has languished, and the vocabulary for talking about why webcomics are compelling (in a way A·I can’t capture) won’t come from an analysis of their technical features. it will come from learning the language of webcomics and understanding why A·I doesn’t (and can’t) speak it.

but this isn’t just true of webcomics. A·I doesn’t (and can’t) speak ANY language. we mostly don’t need to teach people how to recognize A·I writing because it’s kind of obvious. we don't need to say things like “write large paragraphs with long, complex sentences because A·I is fundamentally bad at maintaining a consistent thread thru them”, even tho that may be true. it’s kind of beside the point

so why are we saying things like that about visual art?

@Lady oh. I do think it's interesting to talk about the technical shortcomings of LLM text so maybe we are not aligned on our approach in a more general way

Sign in to participate in the conversation
📟🐱 GlitchCat

A small, community‐oriented Mastodon‐compatible Fediverse (GlitchSoc) instance managed as a joint venture between the cat and KIBI families.