x
When Is a Human Actually "Overseeing" an AI System? — LessWrong