Reaction
Some AI labs may be building models too large for quick reinforcement-learning iteration
A user calls it a mistake for labs to build models they cannot quickly iterate on with reinforcement learning.
TLDR
One post calls a quoted post “absolutely brutal for Mistral.” Replying to it, another user argues that many labs build models too large for them to iterate on quickly with reinforcement learning (RL).
Combined views
37
1 Source, first seen ago