Structure of Intelligence -- Copyright Springer-Verlag 1993

Back to Structure of Intelligence Contents

Appendix 1

Components of the Master Network

Structure: Perceptual Hierarchy

Internal Structure: Composed of a hierarchy of levels, each level     containing a number of processors which may interact with each other and with processors on lower or higher levels.

Internal Processes: Each processor obtains input from processors     on lower levels, and sends output to processors operating on higher levels. Each processor operates by guessing the most likely, simplest "reality" underlying its input, using Bayes' rule and probably, on lower levels, the maximum entropy principle.

        The range of possibilities is given by hard-wiring and/or induction and/or deduction and/or analogy; the a priori background distribution is given by hard-wiring and/or induction and/or deduction and/or analogy; and the properties of potential realities are given by hard-wiring and/or induction and/or deduction and/or analogy. As one goes higher and higher up the hierarchy, induction, analogy and deduction are relied upon more heavily, and the estimation sticks closer and closer to the background distribution. And as induction, analogy and deduction are relied upon more heavily, the processing becomes slower and slower.

        Information is passed from the top down when a processor tells one of the processors that gives it input what sort of things it should look for. This overall regulation operates according to the multilevel optimization methodology.

Dependencies: Induction processor, deduction processor, analogy processor

Functions: Supplies the induction processor and the parameter adaptor with the pattern recognition they require.

Structure: Motor Control Hierarchy

Internal Structures: Composed of a hierarchy of levels, each level containing a number of processors which may interact with each other and with processors on lower or higher levels.

Internal Processes: Each processor gives processors on lower levels goals, and each processor refers to the parameter adaptor for information on how to achieve these goals. The success of each processor is continually relayed to the parameter adaptor.

Dependencies: Parameter adaptor.

Functions: Directly controls the action of the system; the effect of mind on body and therefore world.

Structure: Structurally Associative (Long-term) Memory

Internal Structures: A network of processors, each one containing a certain pattern, which is continually reorganizing itself to approximate the

structure of a network of emergence (i.e. of a network in which X and Y are

stored near all functions which are patterns emergent between X and Y). Any

particular pattern may be stored in a number of different places. Along with each pattern is stored the parameters by which the global optimizer can best recognize patterns in it.

Internal Processes: Reorganization must follow some special-case optimization algorithm. In order for structurally associative organization to be possible, it must be known what functions are patterns in what functions.

Dependencies: Perceptual hierarchy (for pattern recognition).

Functions: Analogy could not exist without it.

Structure: Analogy Processor

Internal Structures: Must contain processors capable of applying the general analogy algorithm to many functions at once.

Internal Processes: Searches for patterns according to general analogy algorithm, which requires numerous searches through structurally associative memory and also, in the case of modeling analogy, special-case global optimization in order to invert certain functions. May determine which types of analogy to use according to analogy.

Dependencies: Structurally associative memory, global optimizer.

Functions: Guides the perceptual hierarchy in its pattern recognition; guides the process of deduction.

Structure: Induction Processor

Internal Structures/Processes: Must be able to continually take in large sets of patterns and operate on them according to rapid, specialized approximation algorithms so as to construct a consistent model of the world. Parameters of these algorithms may be modified by parameter adaptor.

Dependencies: Perceptual hierarchy (for the patterns it takes as input), parameter adaptor.

Functions: Aids the perceptual hierarchy in its recognition of patterns.

Structure: Deduction Processor

Internal Structures/Processes: Must be able to rapidly carry out chains of deductions according to hard-wired or otherwise specified rules. Which chains of deductions are executed is determined by an optimization routine whose initial guesses are determined by analogy. Which sets of rules are used may also be determined, to some extent, by parameter adaptation.

Dependencies: Analogy processor, global optimizer, parameter adaptor.

Functions: Guides the perceptual hierarchy.

Structure: Parameter Adaptor

Internal Structures/Processes: Executes a specific set of optimization problems, some involving pattern recognition, induction and analogy

Dependencies: Induction processor, analogy processor, perceptual hierarchy global optimizer.

Functions: Guides the motor control hierarchy and, to a lesser extent, the deduction, induction and analogy processors. Supplies the structurally associative memory with information.

Structure: Global Optimizer

Internal Structures/Processes: Minimizes and maximizes functions according to some widely applicable strategy. Adapts the particulars of its strategy to the nature of the function in question.

Dependencies: Parameter Adaptor.

Functions: Aids the parameter adaptor, the analogy processor and the deduction processor.

Process: Consciousness

Related Structures: Resides, apparently, fairly high up on the perceptual and motor control hierarchies; ventures into all regions of the mind but never too far from there (e.g. never too deeply into the structurally associative memory).

Functions: According to the quantum theory of consciousness, consciousness has no deterministic effect on mental function; it is rather the means by which the mind participates in the creation of the world.

Phenomenon: Emotion

Related Structures: It is hypothesized that emotion occurs when a processor in the motor control hierarchy which notably fails to reach its goal is also a part of consciousness.

Functions: Emotion, like consciousness, appears to serve no deterministic role in mental function. It is, however, a necessary consequence of significant mental change in any conscious system.