Self-Constructing Computing Systems
Participating Organizations
Institut National de la Sant et de la Recherche Medicale, France (Co-ordinator).
Abstract
As circuits get exponentially smaller and faster, we face exponential increases in their production cost. Current hardware methodologies demand extremely low failure rates for individual components, yet when fabricating huge circuits, yields are still low. Nature has solved these problems. Our neocortex, a cellular computer that generates intelligent behavior, constructs and configures itself starting from a single precursor cell, based on genetic information and interactions with its environment. Understanding this process would revolutionize computer technology. Progress in developmental neuroscience now permits a reverse-engineering approach, abstracting nature's principles into systems of our own design. Here we propose some first steps towards understanding these developmental construction mechanisms so that we can transpose them into novel software design technologies. We will demonstrate, by a fusion of experimental neuroscience, detailed physical simulation, and theoretical analysis, the principles by which a population of real or artificial neurons can grow and assemble themselves into functioning circuits. We will transfer these principles into practice by engineering some initial self-constructing applications, where genetic code is placed in a precursor cell, which initiates a developmental process of cell division, cell migration, neurite growth, synaptic formation, and synaptic adjustment. The interplay between these cellular processes, the gradually increasing self- created cellular environment, and the eventual sensory connection to an external environment, reliably generates the final functional circuit. Current fabrication methodologies do not produce circuits that are capable of self-repair, but our approach yields systems in structural and functional equilibrium, thus inherently capable not only of self-repair, but also of environmentally driven rewiring, allowing the circuits to adapt naturally to changes in input or output devices at run-time.
