Abstract
Biological neural networks engage upon a variety of complex tasks proficiently demonstrated by the wealth, flair and grace of many human undertakings (technological or social). A deeper insight upon the fundamental physical and mathematical architecture can enable reproducing these capabilities in current technological applications. Not all neural dynamic capabilities are desirable; however, there are certain subsects of tasks which are performed with high accuracy, speed and minimal energy consumption such as visual or audio recognition, autonomous driving (or other machinery operations), generative languages and behaviors capable of transmitting information demonstrating varying degrees of intelligence. The objective of this paper is to provide a fundamental approach towards feasibly characterizing relevant neural dynamics whose architectures endows intelligent behavioral operations. Furthermore, this can elucidate how the underlying information processing functions enabling reproducing its desirable characteristics (low power consumption and high reliability) in our own technical operations. Thus, quantifying the microscopic energetic consumption and defining the macroscopic level of information entropy will enable identifying a physical structure that can maximize influence upon the global state of information entropy upon an open system while minimizing the associated energetic consumptions. Enabling such a structure allows maximizing the r information capacity of a system given finite constraints on resources. Therefore, this approach can further provide insight towards generating novel computing architectures which address the current bottlenecks in conventional computing schemes. That is, alleviation memory and CPU interaction bottlenecks which are increasingly becoming apparent as it becomes more and more difficult to keep up the pace of Moore’s Law. The importance is further exemplified by the exponentially increasing computational demands in the recent advent of artificial intelligence.
