Except that LLMs are only being trained to do things that humans can do, not things that humans cannot do.

I have heard a lot of claims like this, where we cannot imagine the benefits of AI because the work will look so different to how it is now, but I have yet to see that actually demonstrated anywhere.