I believe he’s making the (joking) point that since we do not/cannot know what a human would want in the limit of maximial knowledge and reflective coherence (thus CEV), it is not impossible that what we’d want actually IS maximum paperclips.
I believe he’s making the (joking) point that since we do not/cannot know what a human would want in the limit of maximial knowledge and reflective coherence (thus CEV), it is not impossible that what we’d want actually IS maximum paperclips.