No. 119 (00249, 00253)

Family name : ROMANOV
Given name : Sergey
Affiliation : Pavlov Institute of Physiology of RAS, S-Petersburg
Abbreviation :
E-mail address : spromanov@spr.usr.pu.ru
Title : What Do Structures Of Nervous System Do In Contrast To
        Artificial Networks?
Authors : Sergey ROMANOV
Abstract :
     Technology of the neural networks of widely develop for the
decision of "intellectual" tasks in hope to transfer the thought
functions peculiar to the man and alive beings to technical devices.
However, in most cases forget that "intellectuality" of technical
systems is provided with a brain of the programmer, which has forced
"iron" to realize certain algorithms.  Attributing consciousness to
nervous system was by great error in 20 century.  Though "brain" is name
of final anatomic department of the central nervous system, brain as
system of analyzing of signals and of accepting of the decisions,
obviously, is something greater than simply nervous substance.  In this
context all works on saturation by intellect of technical devices and
creation of reasonable machines on only logic structures, which have
appeared as artificial networks, such as the perceptrons, and further as
neurocomputers, are hopeless, though they promote development of methods
of programming in computing medium having the large speed and, in
essence, now in unlimited volumes a memory for simplification of the
decision of tasks that are arising before mankind.  Instilled hopes
about paradigm of the large number of elements for occurrence of thought
process were unfounded.  The use of artificial networks is provided with
a many millions of the programmers thinking out distinct from other
programmers the rules and algorithms of training (of changing of weight
of inputs) of uniform multi-layer structures, which have no the
prototypes to the connections between neurons in alive nature.

     We consider the nervous system as an organ of regulation of all
functions and systems of organism.  The nervous system itself is under
control as of the external signals of environment that are perceived by
organs of sense, so and under direct control by condition of internal
environment of organism, movements of which nervous system adjusts for
providing adequate behavior of organism in space surrounding him.   The
nervous system arises as system of automatic reflex regulation and then
of conditioned reflex regulation, where the principles of
multi-parametric ring regulation provide not only homeostasis of
parameters of internal environment, but also homeostasis of execution of
motor function.  From these positions we showed the structures of spinal
level represent homeostatic (in a terminology of W.R. Ashby) mechanism
of regulation of muscle contractions that transforms bio-mechanically
unstable peripheral apparatus in controlled for supraspinal of
departments of the central nervous system.  The important moment is the
representation about homeostatic regulation of patterns of flows of
impulses in ring structures by properties of elements of these
structures.  In this case, the succession of impulses behaves as a
spring ensuring returning of system to an initial condition after action
of disturbing forces.  The sensory structures extract the dynamic
parameters of signals influencing on organism, and shape an image of an
environment in the central formations of nervous system.  Here according
to a condition of organism the commands of controlling to the executive
organs are forming.  We showed that structure of connections of neurons
of cerebellum and the columns of neurons of cerebral cortex may shape
forecasted   meanings of signals, compensating as delays of conducting
of signals in nervous tracts and as the inertial delays in the executive
organs.  In conformity with a state of central nervous system the
adequate or inadequate perception or recognition of the reality is
provided, and the behavioral acts arise on the basis of feelings.

     Certainly, a wide divergence and convergence of connections between
neurons provide distinguishing of the certain features of biologically
important signals, by carrying out instantaneous classifications as
coincidence of visible and thinkable images, mechanisms of which are not
accessible to artificial networks, though this phenomenon can be
reproduced in virtual space on the screen of monitor.  "Iron" is
considered as a reasonable one only by uninitiated observer.  If we
shall be able to create from "iron" an object that is feeling and taking
a decision according to the own sensation, whether that will be possible
to name this product as the Subject? Where is boundary layers between
neuron network and external and internal environment on which can arise
artificial intelligence?