Making Deep Learning Go Brrrr from First Principles

I can't get past that fist meme. "Pure connectionist models" are scaled up "without inductive biases"?

Did I wake up in an alternate universe again?