Natural, Artificial & Synthetic Intelligence: A Conceptual Discourse

Why do biological systems appear to produce adaptive, intelligent decisions using vastly less energy and computation than modern AI systems? Neural network based Artificial Intelligence can be conceptually described as the interrogable compressed human knowledge that is implemented on vast amounts of computational capabilities. The goal of this paper is to propose a conceptual approach to low computational models for machine-based intelligence. Based on the review of Natural Intelligence and Artificial Intelligence, Mental Activity can be deconstructed into Physical Cognition, Emotional Cognition, Mental Cognition, and Intelligence. A syntax-free Semantic Information which requires a goal-relative semantic concept of information (not the powerful Shannon’s Information for communication theory, compression, and uncertainty reduction) is used to construct knowledge structures that are related by their semantic information. Neither is it Transformers that are dependent on language syntax to infer semantic information. This paper proposes that one reason biological intelligence appears computationally efficient is that it operates through structured semantic knowledge domains rather than through brute-force syntactic or statistical search. Asymmetric Information Resolution (AIR) Models are presented as a candidate architecture for such semantic decision structures. This paper provides a simple Predator-Prey knowledge structure to illustrate the versality of AIR Models. The surprise is that modern AI is based on mathematical theory of networks and not on a theory of knowledge or decision theory, but it works.