Fuck AI

There's so very much in that paper that doesn't seem to suggest what they are saying it does. It suggests people are directly using those tools to create scripts for academic videos instead of their fundamental speech changing. They state that they manually reviewed for "reading vs spontaneous" and found about 30 percent were directly reading a script, but extrapolating the non "reading" samples to not have used AI copy edited outlines in this context is a leap. It would make more sense that they did. These lecture videos were not examples of natural language use in any sense.

Our study is focused on academic communication, yet we anticipate that similar patterns may extend to other communicative contexts.

Seems particularly unfounded (though it really has enough hedges to make it a non statement "similar" "may" with no reference to what context they're thinking of). It's also a preprint that gets most of its models from preprints.

Then the vice article takes every weakness of the paper and actually amplifies them to a really profound degree. We've got researchers trying to push an AI is transformative narrative and a journalist trying to push an AI makes you stupid narrative right off a cliff into "popsci journalists reporting on preprints make stupid claims" pit.