x
Why Optimization-Based Alignment Structurally Cannot Handle Irreversible Cost — LessWrong