Bad ML Abstractions (Generative vs. Discriminative Models)

This is not an abstraction I've really seen in the wild. I understand that GANs and classifiers are seen as different, but I've never seen that as some fundamental division of models. The author's discussion makes sense though.

Personally, I think a better "bad abstraction" to criticize is supervised vs unsupervised. It seems to be a cliche of entry level machine learning questions and had almost no meaning