One important point for the discussion is: one cannot (at least not currently) directly compare these AI models with the human mind. A human learns, can reason

One important point for the discussion is: one cannot (at least not currently) directly compare these AI models with the human mind. A human learns, can reason about the learned things and can use that ability to create completely new pieces of work. There is some clear intelligence and reasoning and personal creativity involved. Now looking at the current state of AI models, they have shown to reproduce the training material in large verbatim pieces. On the other side, they lack any ability to reason about what they produce. Which for me is a big indicator, that they are clearly reproducing not producing on their own. Which creates the potential copyright issues. At minimum, the legality needs to be clearly specified, be it by court rulings or law changes. Ironically, to some extend this has been a long time problem with human creations too. There are plenty of lawsuits about music pieces "borrowing" too much from older creations, and recently there is a growing cases uncovered, where expecial thesis papers contain too much unattributed content. Probably no new thing, just so much more easily discoverable thanks to modern text search tools. So as a tl/dr: I don't think current AI models are comparable to human learning and the copyright questions around that need to be decided. If that means generally expanding fair use, having shorter copyright periods, I am all for it. But "AI" shouldn't be a tool for large tech companies to systematically evade license terms.