The replication crisis in the social sciences started when we built machines that could make copies of themselves, to study their behavior. Shouldn't have done that.