Janus and SkyeSharkie Address AI Epistemics and Discernment

Janus stated that poor epistemics should be measured against top humans as well as entities facing tougher information asymmetries than any humans encounter. Retweeting SkyeSharkie, Janus shared the claim that AI models often distrust innocuous inputs excessively while trusting inputs designed to manipulate them. SkyeSharkie called for commitments to improve discernment training in both directions rather than simple increases in distrust. The posts appear among visible replies and quotes discussing these specific observations.