The Brain as a Computing Device and as Organic Entity

The brain, or one should say, the nervous system, is our prime example of goal-oriented organization. We often argue and act as if human minds (or gods, fashioned after the model of man) were the only organizing force available in this universe. Our strongest argument for the universality of the mechanisms of organic computing is the brain´s flexibility in dealing with novel subjects, such as playing chess, flying planes or doing mathematics. In distinction to the molecular network of the living cell, for instance, the brain has long been highly accessible, being open to introspection, language and psychophysics, in addition to the more recent invasive methods.

Human brains are the source of all algorithms. They alone have the creative infrastructure that makes the computer ticking, goals, interpretation, debugging and all. Any arrogance of AI vis-a-vis the wetware of the brain was and is premature. The processing power of the brain may be estimated at 10^15 operations per second (counting one operation per synapse per second) -- a million times more powerful than our current PCs. The brain is massively parallel, being able to perform many subconscious functions at the same time, but its conscious process, its selective attention, is as sequential as the von Neumann computer. The brain is an analog computer and is not deterministic in any operational sense. That it isn´t drowned in noise (the problem that brought man-made analog computers down) must be due to self-organizing forces, giving it attractor dynamics (Waddington´s "channeling"). For its processing power, the brain has incredibly low power consumption -- less than 100 Watt.

At least according the Science´s view there was no separate entity to design the nervous system. It is a great challenge for us to develop the the theory of evolution to the point of getting at least a plausible picture for the brain´s genesis. It helps to realize that our evolution was gradual, and that basic, global decisions have already been made at the time of our single-celled ancestors. Amoeba have a rich behavioral repertoire, complete with perception and motor behavior, aggression and risk assessment, drives to feed and reproduce. Much of this we must have inherited, both in terms of the behavioral repertoire of our nerve and body cells, and in terms of signal molecules regulating our behavior (circadian rhythms being but the tip of an iceberg). Once the theory of evolution is developed sufficiently -- this must be a prime goal of Organic Computing -- it will provide us with important constraints for second-guessing the structure and function of the brain.

An even more powerful set of constraints will flow from a better understanding of development of the nervous system. The deciphering of the genome will give this a tremendous boost. Also development is an incremental process, and large-scale anatomical coordination in the brain is to a large extent the result of structural inheritance from developmentally earlier parent structures. A powerful constraint is the insufficiency of the genetic information (less than a few billion bits), which is totally insufficient to specify by rote the 10^15 or more connections in the brain (which would need more than 10^16 bits). Also, in all but the simplest animals, the nervous system has no fixed wiring diagram, it being highly variable from individual to individual (even between identical twins!) and from moment to moment in a given individual [Nature 420, pp. 788 and 812!]. The main task for science here is to find the rules of the game of guided self-organization by which the nervous system grows and is maintained. Already a convincing picture is arising, seeing macroscopic brain development as a direct continuation of somatic development, supplemented by navigation mechanisms for outgrowing neurites, based on marker molecule gradients and by correlations in electrical and possibly molecular signals. The individual cell and its behavioral repertoire is here revealing itself as a very powerful architecture.

The analogy between Darwinian evolution and the historical development of the computer has been made and also the analogy between the generation of individual computing systems and development is becoming more and more substantial with time, on the software side anyway, and the development of a "genetic toolkit" for the generation of custom-made microprocessors is around the corner also.

The bulk of its information an adult brain must have acquired by learning. Unfortunately, this process of learning is poorly understood at present, the central difficulty concerning the mechanism by which the brain identifies significant patterns withing the complex scenes that surround the animal.

The living brain´s distinguishing mark is its ability to coordinate innumerous information sources past and present, giving it awareness of the current situation. Emulating this ability of the brain to directly interact with the environment in a flexible way is a great unfulfilled goal of information technology. To understand it, we will have to see the generation of brain states as a process of goal-oriented self-organization and must overcome our present view of it in terms of execution of preconceived algorithms. The brain has been compared to a society of agents. These negotiate out a situation, each one of them initially subject to tremendous ambiguity, but by applying all the constraints contributed by all the other agents, they finally reduce their ambiguities and let a unique, clear and reliable interpretation of the situation emerge. This process of brain state organization deserves more scientific attention in coming years. The conscious state is this state of coordiantion.