x
A Conservative Vision for AI Alignment — LessWrong