arguments opposing generative AI based on the artistic merit of its output are not very useful or actionable because art is simply too subjective to base any legislation on and also most of these arguments devolve into some bullshit about the human soul or the importance of toiling or whatever.
setting aside all the actionable reasons to object to the technology for a moment, I think image generation had a lot of potential for the kind of art where you just kind of see where the medium takes you and curate the most interesting results. the unpredictable and surrealist quality of the output was a lot of the appeal of the early models back when it was just a novelty and we were generating "bloodborne horses" and stuff
it also saves a lot of time when making images of which its presence matters more than its contents anyway, such as background decals in video games or whatever, but that's not really what this post is about.
this "see where the medium takes you" approach is particularly suited to abstract art, and the more abstract the style the harder it becomes to distinguish the computer generated from the human-made. there comes a point where the computer's inability to actually understand what it is depicting becomes a strength as it allows for abstractions that are counterintuitive to a human pattern-seeking brain. this is something where I genuinely think the technology could bring something new and unique to the table.
this medium-first approach to art has existed for as long as unpredictable mediums have existed and curation is also a legitimate and extremely undervalued skill, so self-proclaimed AI artists are not completely full of shit when claiming this stuff takes legitimate skill. the reason a lot of AI generated art that gets posted is mediocre anyway is because a lot of self-proclaimed AI artists have a poor eye for art and are therefore not actually very good curators. a lot of self-proclaimed AI artists are more interested in the speed and quantity of the output than its artistic merit, and because they care little for the merits of artistry to begin with, they lack the eye for art to be able to see the ways in which the image they have curated is kinda ass to anyone who does have a more developed eye for art.
this is how you get people who are deeply wrong about what images are just filler and which ones actually matter too, but that's still not really what this post is about.
I think it's interesting that one of the more common arguments for the effort that goes into AI generated art is that it's difficult to get the model to actually understand your vision. the biggest artistic strength of the medium of generative AI is the unpredictability of its output, so the clearer your vision the less useful the technology is to executing that vision. the computer cannot actually take the image in your brain and make it real unless the image in your brain wasn't that clear to begin with. generative AI in general is strongly marketed as a tool for people with no ideas for this reason, which is very self-damning in my opinion.
I think this is particularly interesting as an argument coming from self-proclaimed AI artists because it shows they don't understand the actual skills they are employing in this process. they do not recognise curation as a skill, they do not recognise the value of having a good eye for art when trying to get worthwhile pieces out of an unpredictable medium. the unpredictability and lack of human understanding is a flaw they have to fight against rather than the whole point of working with this technology. they are wrong about the potential merit of the thing which merits they are trying to defend, which is even more self-damning.
these people with an underdeveloped eye for art are the majority of the actual users of generative image technology, because most people with enough of an eye for art to be able to curate interesting results were getting up in arms about it instead, or simply have too clear a vision for the technology to be of use to them. as a result a lot of that interesting surrealist unpredictability has been trained out of the models to accommodate these majority users that are mistaken about the actual artistic potential of image generation. which is also rather self-damning.
and ultimately it's rather self-damning as well that the most straightforwardly useful application of a tool which people will insist has artistic value is to mass produce images of which the contents don't really matter anyway.