Navigating Conceptual Space; a New Take on Artificial General Intelligence

EasyChair Preprint 6552

Abstract

Edward C. Tolman revolutionized cognitive psychology by proposing a clear distinction between learning and performance in behaviorism. Tolman’s ideas on latent learning and cognitive maps eventually led to what is now known as conceptual space; a paradigm where ideas are represented as locations in a high-dimensional Euclidean space. These insights are easily expanded to consider cognitive navigation between ideas – reasoning – as the basis of intelligence. Here, we explore whether conceptual navigation is plausible by the neoRL architecture, an RL architecture capable of having a distributed state representation as found in the hippocampus. Adopting Kaelbling’s concerns for efficient robot learning to spatial navigation, we test whether neoRL is general across NRES modalities, compositional across considerations of experience, and effective when learning in multiple Euclidean dimensions. We find neoRL learning to be more resemblant of biological learning than of RL in AI, and propose autonomous neoRL navigation of conceptual space as a plausible new path toward artificial general intelligence.

Keyphrases: Machine Intelligence, Neuroscience, autonomous navigation, conceptual space, robot learning

@booklet{EasyChair:6552,
  author    = {Per Roald Leikanger},
  title     = {Navigating Conceptual Space; a New Take on Artificial General Intelligence},
  howpublished = {EasyChair Preprint 6552},
  year      = {EasyChair, 2021}}