It’s good to have a paper like this that fully explicates what is meant when we say a word that the industry has nearly obliterated with overuse. Well done. Are there any current alignment projects incorporating RL or other AI tech into their approaches that you’re excited about? I feel like humans vibing their way to value sets for safety and control may be a more short-lived solution than we thought a year ago.
It’s good to have a paper like this that fully explicates what is meant when we say a word that the industry has nearly obliterated with overuse. Well done. Are there any current alignment projects incorporating RL or other AI tech into their approaches that you’re excited about? I feel like humans vibing their way to value sets for safety and control may be a more short-lived solution than we thought a year ago.