Principles of the self-organizing system

Abstract

Today, the principles of the self-organizing system are known with some completeness, in the sense that no major part of the subject is wholly mysterious. We have a secure base. Today we know extactly what we mean by "machine", by "organization", by "integration", and by "selforganization". We understand these concepts as thoroughly and as rigorously as the mathematician understands "continuity" or "convergence". In these terms we can see today that the artificial generation of dynamic systems with "life" and "intelligence" is not merely simple-it is unavoidable if only the basic requirements are met. These are not carbon, skater, or any other material entities but the persistence, over a long time, of the action of any operator that is both unchanging and single-valued. Every such operator forces the development of its own form of life and intelligence. But will the forms developed be of use to us? Here the situation is dominated by the basic law of requisite variety (and Shannon's Tenth Theorem), which says that the achieving of appropriate selection (to a degree better than chance) is absolutely dependent on the processing of at least that quantity of information. Future work must respect this law, or be marked as futile even before it has started. Finally, I commend as a program for research, the identification of the physical basis of the brain's memory stores. Our knowledge of the brain's functioning is today grossly out of balance. A vast amount is known about how the brain goes from state to state at about millisecond intervals; but when we consider our knowledge of the basis of the important long-term changes we find it to amount, practically, to nothing. I suggest it is time that we made some definite attempt to attack this problem. Surely it is time that the world had one team active in this direction?