One post doubts AI labs will agree on a single independent safety evaluator
The post’s skepticism centers on finding someone with the expertise to evaluate frontier AI, no financial ties to a lab, and the backbone to speak up when things go wrong.
TLDR
A post puts “near zero probability” on AI labs agreeing on a single third-party safety evaluator who meets three conditions: competence to evaluate frontier AI rather than repackage existing evaluations, no financial ties to an existing lab, and a willingness to speak up when things go wrong.
Combined views
61.9K
5 Sources, first seen 18d ago