"Love", here, is my word for something like "being very non-linearly controlled by recurring similiar events". The events are, first, seeing a loved person (or thing?) a few, and then many times; and then associating other events with them: Locations, smells, times, and, finally, thoughts. The "very non-linearity" means that there is no useful and somewhat precise model of the properties of the event (duration, distance, novelty, recentness, participants, ...) and the control effect: It might be that the seventh incident, from far away and at a remote nightly time creates heavy breathing and a huge burst of adrenaline; whereas a straight and near eighth encounter just makes oneself a little happier (whatever sort of control this is meant to be).
In other words: There is no reason that probability, in the sense of "frequency of combinatorial patterns", can be used to understand inductive reasoning, or inductive learning.
An inductively learning system needs to "blush" at, i.e. "exponentially" react to, every encounter of an interesting event, in order to include it into its experiences or storage.
I still assume that the "world" is slowly changing here, i.e., it is "Humean". Is this, in itself, something that must be factored out?
No comments:
Post a Comment