A Brain-Inspired Cognitive Architecture (BICA) Approach to the Neurosymbolic Gap

In this paper we consider a brain-inspired cognitive architecture approach to the neurosymbolic gap. The difference in the abilities of artificial neural networks (e.g., excellent perception) and symbolic systems (e.g., excellent logic) can be referred to as the neurosymbolic gap. Most attempts to combine properties of neural networks and symbolic systems are hybrid combinations of these different systems. A brain-inspired cognitive architecture (BICA), the Causal Cognitive Architecture 5 (CCA5), has both connectionist and symbolic properties. This architecture uses spatial navigation maps as the common data structure and requires spatial and temporal binding of inputs, predictive coding, innate knowledge procedures, and the ability to feed back and re-operate on intermediate results. We show how this BICA approach closes the neurosymbolic gap without the need to overtly combine separate symbolic systems and neural networks. As well, given that the BICA model presented is inspired by the mammalian and in particular the human brain, it provides insight into the mechanisms at work in cognition.