These AI systems will sound the death knell for the creative culture of openness and sharing that emerged during the web 1.0 era with the Creative Commons and OSI-approved licenses and sites like DeviantArt, NewGrounds, and SourceForge, and which still exists today, albeit in a less idealistic form.
Once people realize that by putting something, even just a blog post, out for public consumption, especially if done so under a permissive, share-alike license, they're helping to train an AI that will eventually compete with them for their job, they will not only be less inclined to share it (under a permissive license or not), but be less inclined to produce it at all.
I don't think you actually need copyrighted data to train AI to compete with artists. The argument over copyright is interesting but kind of a sideshow. We already know that copyright protection is not a prerequisite for creativity - as it currently stands copyright only protects the rich.
The real reason why AI is going to kill DA/NG/SF/GH/etc is because of spam.
Let's say, for every piece a 100% human artist can do with no AI assistance, another artist using AI as a tool can make three. They start off with a prompt in DALL-E, SD, or Midjourney, and then tweak the output with prompts and so on to get exactly what they want quicker.
Now imagine a non-artist using the same technology. They don't care about the touchups. They just want to churn out lots of pieces all at once. So that would be nine times more output compared to the unassisted human - three times three. And there are plenty of them doing this, because they all watched the same video on how you can hustle T-shirt and poster stores by submitting lots of AI-generated pieces.
The reason why you are seeing art sites ban AI art is purely a matter of practicality. Many of those spaces are able to provide free promotion to people who post quality content because quality is itself a limited resource. Now the human trying to make their magnum opus - AI or no - is going to be buried under lots of low-effort garbage from zombie modernist grifters.
I think it's fair to say that what this current generation of AI enables is cheaply generating a limitless amount of content of not great but acceptable quality.
And I think you're exactly right that this is going to make it extremely difficult for the great stuff out there to rise to the surface. It's going to be suffocated.
We'll soon see the point where AI can generate text, video, voices simultaneously. All of the tik tok/reels style content people mindlessly scroll all day will be AI generated, delivering money to only the AI's creator, and no one will know the AI is behind it.
We are already seeing the same result of LLMs, with sci-fi magazines (temporarily) no longer accepting submissions and self-publication platforms such as Amazon's being flooded with auto-generated crap.
Hustle culture is part of it, but only one part of what I believe will be a negative outcome for society.
As AI systems display inherent bias, and it is possible to train for that bias (e.g. RightwingGPT), we have essentialy weaponized content generation, and we have no scaleable mechanisms such as fact checking to combat it.
Whether it is low-quality work, spam, disinformation, outright manipulation, copyright violations, or other forms of negative results, we have automated the ability to produce it.
you could argue that this LLM's will benefit more users than the produced value itself lost on sea of SEO garbage. But my point is that the openness wont stop because people will still need to show their craft for better employment opportunities.
Once people realize that by putting something, even just a blog post, out for public consumption, especially if done so under a permissive, share-alike license, they're helping to train an AI that will eventually compete with them for their job, they will not only be less inclined to share it (under a permissive license or not), but be less inclined to produce it at all.