Three Alignment Schemas & Their Problems — LessWrong