This is probably one of the most disingenuous slop peddling comments imaginable.

Papers are written in human language for humans to read.

If humans don't read the papers then you can just let the AI read the code base and unstructured experiment protocol.

What is the point in replicating the human process in this instance? The only conceivable reason is that you want to see meaningless output produced ahead of time.

Why even bother with academic publishing? Just publish the results themselves in a database and let everyone prompt the AI.