Natural abstractions form relatively to their agent
This is an idea that came to me after writing Natural abstractions don’t overlap. How heavily do natural abstractions actually overlap? How should be expect a totally alien cognitive architecture to form natural abstractions?
It seems to me like abstraction based around a perception of time (e.g. an abstraction based on the velocity of a rotating gear) might feel different for an AI which might lack a perception of time or have a different perception of time.