Yeah, this post is groundwork for a better version of “AI cooperation in practice”. See here for more. The original version had a much more complicated proof of a similar theorem, following the steps of Löb’s original proof (aka Eliezer’s picture proof) instead of Kripke’s argument. I didn’t really understand Kripke’s argument back then.
Bounded Löb’s theorem seems to provide an easier way to do what you did in AI Cooperation in Practice. Or are they essentially the same thing?
Yeah, this post is groundwork for a better version of “AI cooperation in practice”. See here for more. The original version had a much more complicated proof of a similar theorem, following the steps of Löb’s original proof (aka Eliezer’s picture proof) instead of Kripke’s argument. I didn’t really understand Kripke’s argument back then.