The practical thing is that AI just speeds up the scientific method, it doesn't replace it. You still have to check if an idea actually matches what happens in the real world. So peer review might change, but the core method still works fine.

Doesn't checking results against reality still matter? My opponent makes a fair point that AI doesn't understand deeply, and I'm not sure how to argue back. I just figure the scientific method is about testing, so it still works.