Hardly a "made-up" conclusion -- just a teaser for the next post, which deals with how we can achieve "extreme generalization" via abstraction and reasoning, and how we can concretely implement those in machine learning models.