x
Why Obedient AI May Be the Real Catastrophe — LessWrong