The problems most people have with generative "AI" were always a result of the likely illegal business models popularized by OpenAI and the cottage industry of thin GPT wrappers that have popped up over the past year.
If OpenAI's usage of this stolen data is ruled not fair use a lot of these issues will go away. There are only two methods of training those models in that sort of world and that's paying the creators or crowd sourcing open source datasets. Either option is a vastly better world.
At least if we can control the data sets we can avoid mis information going in to the training data.
This would also include ensuring peer review for data and a way to remove any sources that have been withdrawn or retracted due to errors or concerns on how the research was conducted.
If writing papers etc people are still going to have to fully cite sources.
