> And remember that academics are always required to do something genuinely new, not just "Uber for ice cream, but with AI".

This hasn't been my experience. A lot of papers coming out today are quite iterative.

They sound exactly like:"what if system X had feature Y". Just with more scientific jargon.