-
There are already good takedowns of this (LRT) but in general, things like Midjourney/GPT/Bard cannot train themselves and are explicitly bad at doing things outside of their training set @60Minutes/1647742247444553732On twitter.com
❤️ 3 Favorites Mood +2 🙂
-
The amount of human labor and compute that goes into training a model is immense. And, every time you want to update a model you have to retrain the whole thing. Training and refining huge models is so expensive OpenAI outsources much of the process to the global southPermalink On twitter.com
