An Expedient Mind: A stimulus-reponse mechanism

AN EXPEDIENT MIND

Tartan Hen Publications : Home | more books | Contact : feedback@tartanhen.co.uk

Layer-1

FIG: The Three component types in Layer 1, and how they are linked.

The Stimulus-response Automaton
In the popular imagination, a robot is a machine which is roughly human-shaped although somewhat angular. It speaks in an electronic nassl tone of voice. It reacts to external events in a predictable way by selecting one of a limited range of responses. Layer-1 is a bit like that. It can't speak of course, and it doesn't really have an identifiable shape. but behavioural psychologists would be able to tabulate the way it operates, by associating every stimulus with an associated response. Do not be deceived by its simplicity. This is the foundation for the whole system. Foundations often look flat and uninteresting. But if a foundation is laid well, it can support some very large and complex structures.

Simple Beginnings: The Flagellate System
Let us begin with the simplest system which I can envisage. The sensory #array is reduced to a single #source. That #source is a light-sensitive dot. There are only three #perceptions, each designed to respond to a particular strength of #signal coming from that dot. The output response #array contains only three elements. These (with apologies to Goldilocks) we will call "TOO HIGH", "TOO LOW" and "JUST RIGHT". What would then have is something akin a flagellate - a speck of pond-life. A real flagellate is a bit more complicated than I have described here, but the simplified version will suffice here for the purposes of illustration. If the red dot sends a #signal to its three #perceptions which is below the one of those trigger levels, the system responds with its "too low" response. It twirls the propeller and drives itself upwards towards the sunlight. If the red dot signal is above another level it does its "too high" response, twirles the propeller the other way, and sinks down into the darkness. The "just right" response stops the propeller. So the organism adjusts its position in the water to place itself at the level where the light is at optimum brightness for it to obtain food.

FIG: This shows a simulation of the simplified flagellate, using the components described above.

The Evolution of Layer-1
From this simple beginning, the system can now be developed in gradual steps to become the more complex system illustrated at the top of this section. All that is required is the multiplication of these components, with a small amount of variation (and specialisation), in the detail. For example, some of the #source" components will be specialised to be sensitive to light of a particular frequency, others to a frequency of sound waves. Elsewhere, some sensors will be sentitive to touch, to the chemistry of the blood, the air breathed, to deformation. As the isolated #sources multiply, they can be organised into #arrays. Some of these arrays will detect the angle of the limb-joints, or be sensitive to muscle fatigue (the concentration of lactic acid). The system, with millions of sensors and perceptions to match, will be able to monitor its environment (inside as well as out). It will be able then to respond with normally appropriate, but standardised actions. I say "normally" because the response will not always be appropriate. Amid the haphazard swings and roundabouts of events, the species will prosper, (and the genetic inheritance preserved) if the responses are appropriate in the majority of cases.
As the system becomes capable of sensing and responding to an ever widening range of circumstances, the priority levels assigned to these various circumstances, become ever more important. They provide the means for the system to select and respond to the most important events. In the example I gave earlier - a clash between the need to feed and the need to flee, fleeing would normally take precedence over feeding. We could summarise that by saying the "fear trumps desire". It is a little early in the development of the system to start talking about emotions, but that expression is a convenient short-hand for the system's behavioural response. It also emphasises the need to separate circumstances into two major categories which I will label NICE and NASTY. A NASTY condition is one which should be avoided, a NICE condition is one which should be repeated whenever possible.
It is very easy, as the paragraph above illustrates, to slip, carelessly, into the language of desire and need. Words like "should" for example, indicate (or seem to indicate) some prior requirement which is separate from the simple operation of the mechanism. But how can a simple mechanism, know what it should, or should, not be doing? It just does what it does. It does what the mechanism is designed to do. So the question we should be asking is, how did it come about that the mechanism operates in a way that gives the avoidance of danger, priority over feeding? The answer, of course, (as is true for just about everything else to do with this system) is that a multiplicity of similar systems tried it both ways, or no way at all, and the only ones which survived to pass on their genetic inheritance, were those which were lucky enough to have the correct set of priorities.

Pseudo-Evolution
We could, if we had endless patience and unlimited resources, simulate that evolutionary mechanism. Manufacture billions of systems and turn them loose to survive or perish. Then manufacture a new generation of systems based on the design of the successful ones. It would not be too difficult to write a computer program, which could read the circuit diagram of an existing system, and reproduced it exactly many times. And as it created each new individual it could inject one or two small random variations into the design. The majority of these variations would, of course, be fatally flawed, but a lucky few would prosper to generate future generations. That's the way evolution works.

Sensory Perceptions
I have described the mechanism of perception only in the most general terms. In an earlier text, I tried to describe how a hierarchical structure of #perceptions could recognise a wide variety of features in the external world. The impression I got from readers of that text, was that either they found it incomprehensible, or, if they were already familiar with these issues, they found it too superficial and not in agreement with neurological evidence. On that second point, I would rely that I am not trying to mimic biological mechanisms. All I am trying to do, is to explain to those who doubt the feasibility of the enterprise, that simple data processing methods can recognise such complex aspects of nature as shape, texture, etc. I eschewed the complicated mathematical methods favoured by current researchers, in favour of more home-spun and understandable methods. Even so it appears that the explanation was rather difficult to follow and tended to divert the reader from the main purpose of the book. I have therefore decided to put all those detailed additional explanations into an appendix (see addenda).
There is one point, however, that I should note here. The evolutionary development of layer-1 does not stop when the higher layers are introduced. Modifications, which might have been unhelpful when layer-1 stood alone, can provide additional survival benefit when the higher layers are able to take advantage. The notes on sensory perception contained in the addendum should therefore be read with an assumption that the whole system is up and running.

The Contribution made by Layer-1
No one would ever claim that the layer-1 system which I have described, (still less the simple flagellum system), as "intelligent". But before we pass on to higher layers of the system we should note this. Layer-1 provides several components vital to the creation of an intelligence system.

(1) It provides the ability to perceive the environment (both inside and out).
(2) It provides the ability to respond to events in many different ways.
(3) It provides a classification scheme of priorities which will enable choices to be made.
(4) It provides a generalised classification of events into NICE and NASTY.

Higher layers are then able to intervene between stimulus and response - to make use of that information, to remember, to select, to analyse and to predict. Those higher layers will, on occasions, overrule the automatic responses of layer-1, and will guide the response in some different direction, better calculated to promote survival. The development of consciousness is just one aspect of the evolution of those higher layers.

Tartan Hen Publications : Home | more books | Contact : feedback@tartanhen.co.uk