Raw LLM Responses

It's contextual. AI art is "good" enough to replace the artist who painted the backdrop scene of a quaint country farm for the organic farms ad in billboards at the public transit station. A lot of stock imagery is perfectly viable to interchange with AI generated content. AI art is almost certainly not "good" enough to replace artists in galleries, comic book/manga two page spreads, or many compositional plates for film or show media. AI Art isn't controllable well enough that you can control the details you need to control to ensure the author's intent is fully captured in the piece. And it likely never will be, because the issue isn't one of artistic talent on behalf of AI, it's one of human language not capturing detail in the same way visual imagery does. The authorial intent of any AI generated piece will always not be an accurate representation of what the artist truly envisioned. At best it might get "kinda close". I do think that most likely we'll eventually come to a point where AI is tolerated when it populates the "unimportant" details in those types of art pieces which ARE precise human expressions. For example, I don't think people really care if the grass texture in a video game is hand painted, a photo, or something AI generated, unless that caring comes from a place of moral prejudice against AI rather from a place of aesthetic preference. It's not supposed to communicate artistic intent, it's just supposed to evoke grass. And I do agree that this will have have substantial impacts on the skillsets of artists which are valued, and the career opportunities available to artists. These impacts will cause many art and creative industry professionals "short term" pain. Some people will experience a very large amount of pain. The cat is not going back in the bag though. Even if artists end up winning a more or less total victory and midjourney and other AI companies have to completely abandon their current top models and start from scratch without using any "stolen" art, even if most of those companies go out of business; the technical lessons are already learned and we now live in a world where industry experts understand how to efficiently build AI models which can generate images. There are MANY companies who have enough imagery that they own, combined with what is public domain, to train a model which can generate pretty good images. Google's image generator projects will barely experience a bump in the road if the content set of deviant art, and all the other photography and art copyrit by private artists etc, was no longer available to their training processes. They have a gagillion photos of random shit in the world taken by cars they've driven literally everywhere. They can't legally use ALL of it, because some of it contains content owned by others. But they can use the vast majority of it. The next generation of models which would appear after such an unilateral court decision would be modestly inferior to current models, and will be especially much less capable of copying the styles of specific popular artists, but they won't be bad enough to dissuade people from using them to generate stock imagery for their powerpoint presentations, or edit the trees in the background of their instagram photos to look cooler. And it will be a very short period of time before AAA game development studios go back to using them for rough concept art and generating grass textures. There no longer is a future available where the artists equivalent to needles and thimbles are not substantially replaced by textile machines. We might delay the impact for a bit, but even that I find quite doubtful. In the long run it never matters how the laborers prefer to work, it only matters what the buyers are willing to pay for. There will always be types of art where the buyers DO want to pay extra for the more hand crafted version of the product. But it's impossible to prevent someone from making art cocacola. And even if Midjourney et al lose on every term in a wildly successful lawsuit on behalf of artists against AI generator companies, there also will DEFINATELY be a black market of old and legally dubious models, which will be no more feasible to suppress than online underground porn. Fuck, the porn industry is already like 50% this type of content. Probably the majority of AI image generation these days is done with open source diffuse models modified by Chinese, Indian, and Russian basement dwellers to imitate cool and popular styles. For an AI artist wanting greater stylistic control, the open platforms are far better at it than the big model providers. There is no one getting paid a use fee or being informed of the violation when I run stable diffusion on the GPU in my computer, or even Amazon's cloud GPUs for that matter. It's simply not technically possible, and probably highly illegal, for them to log all the content being passed to those types of cloud services. Even less so for the thousands of Chinese knockoff models. The internet will forever be full of images generated by these content creators, most of which make absolutely zero money off their creations, and pay no one in the process of creating their content, and thus there's no real way to attack their pocketbooks or suppliers like you could do with a company like midjourney. It's going to be much more difficult to legislate against companies who train models on grey market, zero copyright, images. While those models may not be able to understand "in the style of studio ghibli", they will be capable of generating images as visually impressive as those styles. An up and coming artist may not have to worry too strongly about their style being copied by an AI, but they will still be competing with those AI for jobs where the art is intended to be used as stock imagery or any type of low bandwidth imagery (things that need to "look good", and don't need to say anything of artistic merit).