DanielFilan comments on AMA: Paul Christiano, alignment researcher

[DanielFilan]

What is the most common wrong research-relevant intuition among AI alignment researchers?