Leveling Up: Strategies to Achieve Integrated Cognitive Architectures

Human-level cognition (most uniquely characterized by our abilities to use language) should be seen as a superset of functional and behavioral capabilities shared by lower life-forms including animals and insects, and this perspective ought to principally guide our strategies for developing integrated cognitive architectures. Just as the study of biological model organisms has led to tremendous advances in our scientific knowledge of genetics and cellular function, the study of embodied cognition in simple agent-environment simulations can yield similar advances in Cognitive Science, Artificial Intelligence, and Robotics. By working first on the foundations of intelligent interaction with one’s environment, and by focusing on core functions such as predictive and inductive learning, probabilistic goal-directed behavior compilation, and empathetic reasoning, we can better establish the grounding that the physical symbol system hypothesis assumes (Newell and Simon 1976), yet often without explicit demonstration of a mechanism to derive symbolic relations and semantics from raw sensory data. Logic and language are seen to emerge from our willingness to make discrete simplifying assumptions in a continuous and probabilistic world of experience, and developing a Standard Model of the Mind can help build much-needed bridges between historically non-aligned research communities.