>For starters, he says, researchers should conduct "ablation studies" like those done with the translation algorithm: deleting parts of an algorithm one at a time to see the function of each component. He calls for "sliced analysis," in which an algorithm's performance is analyzed in detail to see how improvement in some areas might have a cost elsewhere.

...wait a minute. ML researchers don't already do this?!

Forgive me if I come across as arrogant or naive; I am a scientist but not in the ML field. I just can't fathom ANY field of research in which this very basic approach isn't followed.