Davidad Questions Ghelich View on AI Formal Verification

David Dalrymple posted a comment disagreeing with Ghelich about formal verification in AI. He argues that formal verification extends beyond basic checks to include comparison of explanations through imprecise Bayesian hypothesis comparison. Dalrymple also suggests that an explanation market could help align incentives in the process. At the same time he notes that humans in a trusted quorum would need to endorse the chosen axioms along with the outcome specifications. The statement comes from his account where he discusses technical alignment and provably safe AI systems. The packet presents this as his direct quote on the topic.