Computational Virtuality as a Form of Artificial Intelligence

Virtual environments give us the opportunity to understand perception and action in ways that are not readily available in everyday life. One of the most surprising aspects of behavior in virtual worlds are so-called cognitive gaps. Cognitive gaps result from the controlled introduction of latencies and deformations into otherwise naturalistic settings, and have surprising implications for brain and behavior. These cognitive gaps come in three flavors: misalignments of the sensorimotor loop, changes in the timing of sensory accommodation, and the decoupling of multisensory integration. This leads to a set of cognitive effects called virtuality, which are theorized to form a functional brain network representing a diverse group of functional areas. Virtuality can be explored further through the use of computational agents. Our agent's sole imperative is to experience and overcome cognitive gaps, the ability to do so results in an agent exhibiting something we identify as physical intelligence. Physical intelligence emerges from five basic cognitive functions: two attentional (integrate and adjacency) and three sensorimotor (displace, track, and sample). In this talk, we will explore the architecture of these agents, some preliminary experiments, and implications for our understanding of the development and evolution of animal cognition.