Stahl67@ bmp

204 WALTER R. STAHL

with the leaming of imitative behavior (Rosenbaum and Tucker, 1961) and a report on the responses of a bobwhite quail to a living hawk and various artificial physical models of hawks.

This section has dealt with simulation of circumscribed thought algorithms and simple leaming processes in humans and animals. In addition, one may simulate sełf-organization and complex adaptation, which is the subject of Section X and clearly in the domain of algorithmic modeling. Several reports are available on brain models with these higher capabilities, such as those by Uhr and Vossler (1961), in which modę of learning behavior may be varied, and Kapłan and Sklansky (1965) on complex stochastic models of behavior, in which prior experi-ence alters the futurę probabilities of subseąuent reactions in a complex manner.

As was madę elear in the pioneering work of Turing, there need be no metaphysical problem in simulating human thought or intelligence. The problem is rather one of recognizing and clearly stating what particular kinds of brain function are of interest and how similarity of the model and prototype are to be determined in an objective manner, i.e., which specific performance criteria (test algorithm) govern the related systems.

X. Self-Organizing Systems and Bionics

These two terms are often lumped together, but in reality usually deal with rather different kinds of modeling attempts. Bionics may be defined as the technology of creating operational models of portions of living systems, usually to serve some practical need, freąuently military in naturę. Bionics models may be physical, electrical, or molecular analogs, but are not usually expected to consist simply of eąuations or Computer algorithms. Self-organizing systems, on the other hand, are with few exceptions simulated as Computer algorithms, and moreover are very abstract ones. Most commonly the term has been used in con-nection with models of memory or brain orgainzation, but it also applies to social phenomena, cellular self-reproduction and tissue differentiation.

A. Self-Organizing System Models

The similarity invariants most applicable to self-organizing systems are again probably of the performance-score type, although various specific numerical measures of organization, particularly those based on information theory may be applicable. A very generał approach was suggested by Ashby (1956), who defined a completely nonspecific “black box” that switches among possible “states” as the result of particular