Clarifying inner alignment terminology — LessWrong