Current models won't write anything new, they are "just" great at matching, qualifying, and copying patterns. They bring a lot of value right now, but there is no creativity.
You can only complain about creativity if you were actually being creative. 99.99999% of the industry was not.
But sure, for the 0.000001% of the industry coming up with new deep learning algorithms instead of just TF/PyTorch monkey-ing, maybe the LLMs won’t help as much as a good foundation in some pretty esoteric mathematics.