Post Digital Computation

Ron Cottam, Nils Langloh, Willy Ranson & Roger Vounckx

Abstract

            Careful reappraisal of computation enables us to derive a model for a processing machine which is closer to a generalised description of the universe than is its precursor of binary logic. We describe the preliminary investigative formulation of an environmental-reaction processor of this kind, and compare its properties to those associated with naturally occurring phenomena and living systems.

Keywords

            Natural computation, Query-reflection, Symbolic models, Non-formal logic

Introduction

             "Road-maps" currently in use by the microelectronics community run out somewhere between the years 2005 and 2015. It is difficult to predict beyond this point how the recent explosive expansion in digital computer technology can be maintained using currently predicted devices and fabrication methods. In this scenario the continued reduction in device size required to accommodate progressively more complicated computational circuits on available chip or wafer surface areas, along with the associated increasing difficulty in correctly predicting and testing the operation of resulting systems, will combine to limit performance expansion. It becomes questionable as to whether we can continue to insist on rigidly deterministic and therefore perfectly predictable computation techniques comparable to those currently implemented in digital circuitry, or whether we should examine the possibilities of more context dependent methods, not only in terms of abstraction of the computation itself, but also where possible by using problematic aspects of ULSI digital design to our advantage.

            We conclude that not only does a move in this direction enable us to avoid some of the major difficulties of purely digital systems, but that in doing so we find ourselves approaching computational techniques which more closely resemble those encountered in the living world, and which are characteristic of the generalised description of the universe as a coherent entity.

Background

             The reappraisal of computation implies not only rethinking how we use machines, but also questioning of our current understanding of the nature and applicability of the machines themselves. To more satisfactorily approach this evaluation we start by considering the validity of a number of presupposedly "correct" ideas behind our understanding of computation. Following on from the suggestions of David Bohm, we first attribute a coherent nature to all elements of our surroundings, by describing them in terms of agent-generated capability rather than independent properties. This leads us to a description of the character and importance of scale in information processing, and consequently to question the use of serial or parallel processing techniques. We draw a comparison between the occurrence of meta-states associated with the stability of natural organisms and the intermediate evaluable results which appear in the course of long computations.

            We find that a novel form of environmental-reaction processor which combines the advantages of both serial and parallel processing corresponds well with general criteria which we can lay down for a survival-conscious computational entity, and the processor itself forms the basis for an even more general symbolic model.

Environmental-Reaction Processing

            Having set down the requirements for computation capable of supporting survival in an environment exhibiting the evolutionary generation of unexpected events, we must now approach the task of implementing these ideas into a preliminary model for a practical environmental-reaction machine. We need to communicatively integrate in some way distributed conservative processing and reductive decision-making, and realise the selection of representative models in a manner which takes account of the relationship between their computability and the variability of the time-scale within which their application must be effected.

            We find a configuration, acronymically referred to as "Aquarium", which can be manipulated to not only connect these two computational forms together but ultimately to absorb them into itself. We can achieve data reduction through the application of an assembly of different models, all of which correspond with some degree of accuracy to the empirical characteristics of the environment in which the computation is to operate, and all of which are therefore equivalent to collections of post-imposed rules whose existences depend not only on the prior accumulation of experimental information, but also on the execution of a large measure of processing at some time in the past. In the assembled scheme this priior processing serves not onl to establish the individual model descriptions, but also to set up their implantation into a sequentially accessible configuration. Far from being a complicating disadvantage, this method ensures that most of the computation required to establish a suitable stimulus-reaction has already been carried out in an earlier time-frame, where it has no influence on the currently available reaction rate.

            The resultant arrangement implements multiple reflections of an incoming environmental query, which provides continuously updating response to the requirements of a possibly time-limited external context. Prior definition of the optimal depth of query-penetration into the processing assembly is not required; it is sufficient to wait as long as is permissible in the given context, and the "then-current" reply is the best we can obtain in the time which was available.

Conclusion

            The processor we describe is based on a careful evaluation of the character of naturally-occurring information processing systems, from the use we make of our own brains and the way they react to external stimuli to the mechanisms involved in the inherent processing which leads to a universal coherence in our environment. Further evolution of the design depends on the implementation of modified forms of the mathematical treatment of chance and probability, to establish a formal basis for dealing with the non-formal rationality implied by intuition. Aquarium itself forms a symbolic only-partially-deterministic model for rationality, which can be used to replace its predecessor in a way which is far beyond the adoption of fuzzy logic.
______________________________________________
______________________________________________