I thought this was a good post! In fact, I had just read a CVPR '18 paper that did the kind of thing he mentioned--presented an 'interpretable neural network'

I thought this was a good post! In fact, I had just read a CVPR '18 paper that did the kind of thing he mentioned--presented an 'interpretable neural network' that assumed what interpretability meant...