List #3: Why not to assume on prior that AGI-alignment workarounds are available — LessWrong