Hacker News new | past | comments | ask | show | jobs | submit
juancn | on: Inside the M4 Apple Neural Engine, Part 1: Reverse Engineering
In principle most if not all inference hardware should be usable for training.

Efficiency is the question.