In this paper we analyze our current understanding of genes, neurons and the neocortex and draw a parallel to current implementations of cognitive computing in Silicon. We argue that current information technologies have evolved from the original stored program control architecture implementing the Turing machines which allowed us to model, configure, monitor and control any physical system using a Universal Turing Machine model. However, large scale of distributed computations and their tolerance requirement to fluctuations have created new challenges. We suggest that a new agent based computing model extension to the current Turing machine meets this challenge and provides a path to cognitive self-learning and self-managing systems.