Discover more from Zhebrak et al.
We make decisions relying on learned representations and human priors. Civilization, with its laws and narratives, is an intelligent system with a set of constantly updating abstractions. By learning, we revise our internal hierarchy: create, replace, modify, and compress acquired concepts. We naturally seek to adopt clear and compelling abstractions that could be seamlessly integrated into the internal topology. These are also the ones easier to promote. With inherent limits to the speed of processing and capacity for new data, we make use of external devices and systems to outsource computations.
While representations serve as a map of the universe, values manifest themselves as rules and procedures for representation management and navigation. Systemic problems can be rendered as abstraction hierarchy issues: wrong maps lead to distortions and dysfunctional behavior, gaps in abstractions produce fractured information flow, and value mismatch impels crisis and conflict. Intelligence can be associated with the ability to operate an internal abstraction hierarchy, carry out inference calls, and turn them into actions. Efficient representations allow for better search and navigation and can be re-used and easily updated as they grow obsolete.