x
Shaping the exploration of the motivation-space matters for AI safety — LessWrong