Pitching the Redwood Research substack: We write a lot of content about technical AI safety focused on AI control.[1] This ranges from stuff like “what are the returns to compute/algorithms once AIs already beat top human experts” to “Making deals with early schemers” to “Comparing risk from internally-deployed AI to insider and outsider threats from humans”.
We also cross post this to LessWrong, but subscribing on substack is an easy to guarantee you see our content.
Pitching the Redwood Research substack: We write a lot of content about technical AI safety focused on AI control.[1] This ranges from stuff like “what are the returns to compute/algorithms once AIs already beat top human experts” to “Making deals with early schemers” to “Comparing risk from internally-deployed AI to insider and outsider threats from humans”.
We also cross post this to LessWrong, but subscribing on substack is an easy to guarantee you see our content.