I agree, but only in the sense that I think any amount of copyright protection for scientific papers is absolutely absurd. The creativity involved in papers is minimal and a good chunk of that research is funded by the government, so paywalling it is criminally unethical.

Also, if we're going to bin the entire concept of copyright, can we at least be equal about it? I'd rather not live in a world where humans labor for the remnants of their culture in the content mines while clankers[0] feast on an endless stream of training data.

[0] Fake racial slur for robots or other AI systems.

I agree. I think that copyright should be abolished entirely, especially for scientific articles (if they are good quality scientific research then I think they would be too important to be copyrighted, in addition to the other stuff you mention), but also for anything else too.

Nevertheless I thin there is another thing against the LLM training, which is that the scraping seems to be excessive (although it could be made less excessive; there are many ways to help with making it less excessive) and I think it requires too much power (although I don't really know a lot about it).

These are two separate issues, though.