The basic Bayesian strategy for thinking about the world (assign credences to every possibility, update when new information comes in) even works for purely logical statements, not just empirical ones. "The first fully rational beings will be AIs."http://marginalrevolution.com/marginalrevolution/2017/11/logical-thoughtefficient-markets.html …
-
do you still have that link to the OpenPhil reviews?
http://files.openphilanthropy.org/files/Grants/MIRI/consolidated_public_reviews.pdf … (papers 3 and 4 there are predecessors to the LI paper)