My own specialty is A.I., but I figure this is an essentially philosophical topic. The existence of patterns and of change, or, said differently, of continuous change at multiple levels of abstraction, in space or in time, seems (to me) to be the basic prerequisite underlying much of machine learning. Without patterns or without change, there is nothing to learn and no reason to learn. This of course leads to questions such as: why are there patterns rather than none (e.g. galaxy clusters, galaxies, star systems, forests, species, communities; re-usable skills, generalization between images, etc.); what's a pattern anyway; is this a characteristic of the world or our perception; etc.
I'm looking for work that focuses on these specific topics. That might include philosophical work, but also work in mathematics or physics or information theory that formalizes the issue satisfactorily. I figured philosophy was the most general starting point, and I'd like to find very general thinking on the issue, hence my asking this question here.
As far as I know, philosophy of AI is largely still "stuck" on completely different questions that stem from a logicist approach to A.I.; and reflexions based on connectionism do not seem to be nearly as sophisticated.