[MathYouF]

This is a very important point about shifting the calculus of automating human labor when doing it incompletely. There are still big debates in the self driving car world between people who want to get to level 2-3 (partial/conditional automation) vs level 5 (full automation).

The earlier group often says since Edge cases that they can't automate now constitute only <5% of training scenarios they encounter, they've automated 95% of the job. But with what you're saying, we can't really expect the calculus to work that way.