Yup.

I do tend to think this is the only way it could have gone. Either this, or figure out some form of attribution and remuneration.

I'm glad Stability.ai has stepped up and taken a real position on this (I also happen to agree that it's the right one), rather than waiting for litigation to force their hand. And, to be fair, the actual image data is still in there, so the model's themselves aren't any less capable. To me, that indicates an effort to maintain the quality and capacity of the models without explicitly stealing from contributing artists and, more importantly, proposes a state of the industry where training the models on copyrighted data is legal, so long as their usage protects artists from stylistic infringement. It's a hugely significant position, which is actually very much in support of the AI community. The other option would be for copyrighted images to be completely banned from all datasets, which would leave the models themselves much less capable and would be much more detrimental to the technology.

Simply put; generative AI can't do what it does without the human art community. So we need to respect those who make the field possible.