Hacker News new | past | comments | ask | show | jobs | submit login
Deep learning has a size problem (fritz.ai)
3 points by nnx 12 days ago | hide | past | web | favorite | 1 comment

This article confused me, the author referenced the time and power required to train a model, which granted is lots, but then talks about phones and mircontrollers which are less powerful but are far more numerous than the machines currently being used to develop these models.

Isnt that the point? The powerful machines create the models through training/data processing, then those models are loaded onto the less powerful units for use?

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact