In a post-scarcity solarpunk future, I could imagine some reasonable uses, but that’s not the world we’re living in yet.
AI art has already poisoned the creative environment. I commissioned an artist for my latest solarpunk novel, and they used AI without telling me. I had to scrap that illustration. Then the next person I tried to hire claimed they could do the work without AI but in fact they could not.
All that is to say, fuck generative AI and fuck capitalism!
No, people are saying that if you mass scrape art from the internet that you don’t hold the copyright to in order to create an image generator that you then turn around and try to sell access to, you’re violating the copyright of those artists (on top of being an incredibly unethical douchebag).
If the artwork they’re using to train the algorithm wasn’t valuable then they wouldn’t be fighting tooth and nail in court to be allowed to do whatever they want with it. They’d just shrug, say okay, and use whatever copyright free stuff they had at hand. If they didn’t need it then they wouldn’t do it, and if they need it then the people whose labor its very existence depends on should get a slice of the pie.
I’m not a fan of copyright in general, but I’m not sold on there being any ethical issue with scraping images to produce training data. People can cry “Copyright infringement!” if someone is using a machine learning model to produce something that’s recognizably derivative of specific work present in the training data. However, I don’t think it’s appropriate in most cases, as the output is often transformative. Also, if you want to go down the intellectual property rabbit hole, a lot of art websites put in the ToS that works could be sold as training data by the controlling entity of the website (at least until people got up in arms about it in late 2022/early 2023).
TL;DR: In my opinion, the output is too far removed from the input to warrant people from getting a slice of the pie, and most people didn’t have any basis for a legal argument until about two years ago.
Beginning artists also need good reference material to become good artists and create new transformative material, is that also copyright infringement? For training to be useful you need more refined material than what you can currently produce, that’s just how knowledge works. The goal of these AI isn’t to produce the same as it’s reference material, if it was then you’d have a case. You can easily see from the output of these generators that the vast majority of what it produces is transformative, confirming it’s intended goal.
Scraping data is also very well established as not infringing on copyright if used for analysis purposes. And if you’ve ever done any kind of analytical research yourself for a PhD or any kind of higher educational degree you know this to be a fundamental freedom required for a healthy society, not even just for artists to learn.
Proposing it should be seen the way you put it would essentially turn ideas into a property one can own and license, and I can tell you now, the same companies you probably dislike will own so many of these ideas that you could effectively do nothing without paying a license to one of them. Is this what you want?
And well, shouldn’t need to be said but, if a company gets sued when they think they’re in the right, they’re going to defend themselves lol. And as far as I know none of these lawsuits have been settled in the favor of artists claiming copyright infringement.