avatar
JG
266 Twitter followers
Follow
Founder & CEO @Evertas | Security, Privacy, Risk, Blockchain, Crypto | Not providing investment advice; opinions are my own; retweets are not endorsement
Posts
avatar
JG
03-29
1. This might mean that Ilya was wrong when he predicted that we’re “moving from scaling the recipe (pre-training) to scaling the thought process (inference)”, if training at scale resulted in a dramatic improvement 2. The costs for other labs to replicate it, means the present is probably the most access most of the public will have to the most advanced models. From this point the cost will diverge into K curves, where the average consumer uses open source level models.
Andrew Curran
@AndrewCurran_
03-29
Three weeks ago there were rumors that one of the labs had completed its largest ever successful training run, and that the model that emerged from it performed far above both internal expectations and what people assumed the scaling laws would predict. At the time these were
CRV
1.18%
loading indicator
Loading..