The problem with enshrining every model in its own hardware is that when – inevitably – models *change*, the hardware is “stuck” wherever it was crystalized Potential benefits, too: “I want a Flurglepuff 1.7b chip – not 1.7a or 2.1g. Thank you” https://t.co/Hm58JBf5wU https://t.co/jG89c2DjmY

The problem with enshrining every model in its own hardware is that when – inevitably – models *change*, the hardware is “stuck” wherever it was crystalized

Potential benefits, too: “I want a Flurglepuff 1.7b chip – not 1.7a or 2.1g. Thank you”

https://t.co/Hm58JBf5wU https://t.co/jG89c2DjmY

— Warren Myers 🐧🐿 (@warrenmyers)
Mar 6, 2024

from Twitter https://twitter.com/warrenmyers

March 06, 2024 at 11:23AM
via IFTTT