Neural Model-Applying Network (Neuman): A New Basis for Computational Cognition

Abstract

NeuMAN represents a new model for computational cognition synthesizing important results across AI, psychology, and neuroscience. NeuMAN is based on three important ideas: (1) neural mechanisms perform all requirements for intelligence without symbolic reasoning on finite sets, thus avoiding exponential matching algorithms; (2) the network reinforces hierarchical abstraction and composition for sensing and acting; and (3) the network uses learned sequences within contextual frames to make predictions, minimize reactions to expected events, and increase responsiveness to high-value information. These systems exhibit both automatic and deliberate processes. NeuMAN accords with a wide variety of findings in neural and cognitive science and will supersede symbolic reasoning as a foundation for AI and as a model of human intelligence. It will likely become the principal mechanism for engineering intelligent systems.