因果推断的本质是超越数据,询问如果我们干预会发生什么。
Train people well enough so they can leave, treat them well enough so they don't want to.
Robots should be partners, not opponents, to humans.