DanielFilan comments on AXRP Episode 4 - Risks from Learned Optimization with Evan Hubinger