Honestly I'm kind of surprised how little buzz there is about the A12's neural engine. ~5 tera-ops in your pocket, and it can run your models. For this particul

Honestly I'm kind of surprised how little buzz there is about the A12's neural engine. ~5 tera-ops in your pocket, and it can run your models. For this particular set of math, that's comparable to a top-of-the-line desktop GPU from 3 years ago.