The development of metabotropic neurophysiology has given Nature additional neural capabilities which extend categorically beyond the uni-dimensional ionotropic behaviors of basic phasic and tonic assemblies, new functionalities that could not be accomplished with ionotropic signaling alone, and capabilities that solved the dilemma of the dangling photonic direction vector by bringing synthetic emotive signaling to our motile metazoan species, in addition to providing the neural potential to form the organic behaviors of perceptual diffraction and selective attention.
Each of these potential behaviors could provide increased survival value as adaptations individually, but collectively they could also provide the substrate for ever more neural sophistication in sensory response as Nature forms the feedback mechanism between the emotive complex and the photonic sensory apparatus that might promote their exploitation.
But just as Nature had to side-track the grander evolutions of ramified antennae in order to develop non-contact exteroception, the evolutionary possibilities of perceptual diffraction and selective attention would also have to be subordinated before Nature could fully exploit their potentials, as that harsh taskmaster, the environment, intervenes with another lesson for Natures’ designs.
As the dialog detailed in the development of photonic exteroception in discussion 9F, the direction vector being signaled by this modality was itself just a geometric composite of the individual shadow directions signaled by the unitary ommatidium of the organisms’ compound eyes. And this singular composite vector scheme worked fine as long as there was only one shadow in the visual field of the organism. But when there was more than one discrete shadow in the overall visual field of the organisms’ photonic apparatus, the summation network produced a composite vector that in most cases pointed to no particular shadow at all. Now, as the geometric average of all of the shadows in the visual field, the composite vector still had some organic survival value as Nature evolved the processes to synthetically determine their emotive characteristic. But Nature would have no time after that evolution to deal with the apparent nonlinearity of signaling in any environment which presented multiple shadows.
The bottom-up engineer should recall that each ommatidium in the photonic apparatus of an organisms compound eye did indeed develop an individual, albeit geometrically coarse, direction signal for those shadows as sensed at that particular matrix position of the compound array. It was the slight differences in a particular shadows’ orientation over all of the ommatidium in the array that the neural assemblies of the organisms’ sensory complex summed, in a simple summation network that literally averaged all of the ommatidium direction vectors, forming a singular composite direction vector.
The synthetic emotive characteristic of shadows in Natures new metabotropic synthesis was based on changes in that composite vector from the photonic apparatus, but in an environment of multiple shadows where, say, one shadow moved and another did not, the composite direction vector to their geometric average would provide a false signal to the metabotropic processes that are abstracting any change. Yet again, the environment provides selection pressures for adaptations to its demanding nonlinear lessons.
The apparent nonlinearities produced in this particular situation are not a product of the synthetic abstraction machinery, but come about from the geometric averaging of the composite vector. Since the summation network which averaged all of the individual ommatidium signaling had been serviceable for environments in which there was only one discrete shadow, the real engineering adaptation that Nature needed was the spatial discrimination between individual shadows before the calculation was made for any relative direction vector.
With the array of multiple ommatidia in the primordial compound eye, the geometric “mapping” of the visual field was essentially performed through the physical configuration of the compound array itself. But the individual ommatidium of the compound eye could not perform any discrimination between discrete shadows themselves, so the “mapping” of the visual field would have to be brought into the sensory complex of the organisms’ neural array, in order to make a neural apprehension of shadow segregation.
From an engineering perspective, the individual ommatidium elements of the compound eye could not perform shadow discrimination because each ommatidium had not developed the capacity to compare its signaling with neighboring ommatidia, they did not contrast their signaling with their neighbors.
The individual ommatidia in our metazoans’ compound eyes had not developed the neural ability to contrast their photonic signaling with that of their neighbors not because that adaptation was outside the capabilities of Natures’ designs, but because the development of local contrast processes between ommatidia also required a global, array wide step, in which the local contrast values could be associated in a manner that would resolve a global contrast boundary map, which could then determine the discrimination between shadows as a whole. Because this faculty was beyond the abilities of Natures’ current toolkit of neural assemblies, and the architecture of the compound eye, there became a need for a new neural structure, one that would define a new form of sensory signaling of a type that moved from local phasic activity to global associations.
This necessary second step of global associations highlights the inadequacy of defining neural activity solely by its connective dispersal. Defining neural activity by connections alone cannot express the functional concept of phasic processes among non-interconnected unit coalitions. Defining neural processes merely according to their synaptic arrangement alone cannot express the definition of contrast as a state, which must be looked at as neurophysiological signaling that has temporal as well as phasic components.
This inadequacy is also found in artificial neural networks, which typically model this contrast mapping process through convolutional arrays and pooling layers. These models have demonstrated encouraging success with pattern matching against a closed set of target patterns, but their inadequacy emerges when the models are used for open generalizations, generalizations from which global associations can emerge, allowing our evolving organisms to evolve from merely reacting to their environment to developing cognitive abilities regarding that environment. Even the use of recurrent neural architecture, oftentimes used in neural network research to emulate the attention mechanism, cannot bridge the functionality required to produce the global associations needed for open generalizations.
While implementing a composite direction vector, the neural array of our motile metazoan provided the sensory apprehension of photonic signaling by the calculation of a generality across the entire population of exteroceptive signaling. But this calculation was reliant on the proviso that there existed only a single constraint in the variety of that signaling.
One of the most fundamental concepts in cybernetics is that of difference, and developing a conceptualization of change versus the basic idea of invariance. And as we observe the interplay between the opposing concepts of change and invariance in our human observations of the environment, we discover another important concept of cybernetics, that of the relation of constraint.
Constraint is a relation between two sets, or possibilities, each set representing the possible conditions or states of an environment we are observing. Constraint occurs when the variety of states that exist under one condition is less than the variety that exists under another. Here, the “variety” of the larger set is composed of what might happen if the behavior of the environment we were observing was free and chaotic, and the variety of the smaller set is composed of what actually does happen.
Constraints indicate that some variance in a system is not independent. For example, donning our cyberneticist hats, we might observe a metal box which has two shafts protruding from it, and we desire to deduce something of its inner workings. If we were to observe both shafts as they turn at various speeds and in various directions, we might be able to better elucidate the mechanisms driving their motion. And if we were to observe, say, the motion history of one shaft, and then observe that the motion history of the second shaft is wholly indifferent to the motion of the first shaft, we might hypothesize that the shafts are driven by separate mechanisms. But if we were to observe that as one shaft changes motion, in speed or direction, the other shaft was also to change its motion, this might indicate that the shafts are not free to operate independently. We would hypothesize that a constraint existed, that the motion of both shafts is somehow constrained.
Cybernetics theory goes on to propose that the existence of any invariant over a set of phenomena implies a constraint, for the existence of the invariant implies that the full range of variety in behavior does not occur. Further, as every law of nature implies the existence of an invariant, it follows that every law of nature is a constraint.
Following this, the bottom-up engineer can perhaps see that one of the top level goals of his artificially intelligent agent should be the determination of those constraints, the laws of nature, that exist in the environment of that agent. But we have also seen how wily the environment can be in dressing up these constraints with the camouflage of multi-dimensionality.
So, returning to the evolutionary sequences occurring in our primordial metazoan species, it would be instructive to see how the adaptations to resolve individual shadows in the organisms’ photonic sensation will represent Natures’ first resolution of constraints in the environment, lessons that can serve to define some of the concepts that the bottom-up engineer will need to form in the engineering of cognition in artificial agents. Of course, the adaptations in our current imagination scenario are moving the organism species toward what the Organon Sutra has defined as the pre-cognitive stage, but before there is an understanding of cognition, the bottom-up engineer must understand those conceptualizations leading to cognition that must be present in the pre-cognitive stage.
And the very first part of these conceptualizations is an understanding of the creation of a uniquely different neural structure, one which can provide the new functionality needed to move from local phasic activity to the global associations of sensory signaling. And that structure is referred to as the Neural Map.
In the photonic sensation of our primordial metazoan, specific volumes of space in its visual field are transduced by the individual ommatidium of each compound eye, and the overall topology of adjacent volumes in that space creates, in effect, a “sensory surface”, which is maintained in the geometric array of the organisms’ compound eyes themselves.
A neural map is created with the ordered projection of that sensory surface onto a mosaic structure of neural elements within the organisms’ sensory complex, elements whose mosaic arrangement and interconnections preserves the topological geometries of that sensory surface.
Neural maps will constitute another utility assembly structure, in fact a most important and central one, that Nature will reuse over and over, in building block fashion, as she progresses toward the neural architectures that will express the behaviors of cognition, and ultimately, of intelligence. And the Organon Sutra will demonstrate to the bottom-up engineer how to translate these building block assemblies and others into the engineering mechanisms that will constitute the artificial agent of a particular design.
But a basic connectionist implementation of neural maps, with their attendant components of uni-dimensional phasic and tonic ensembles, cannot by themselves pierce the veil of dimensionality that the environment fashions to hide the constraints our agent will seek to discover. Although the discrete elements in a neural map have a physical domain relation to one another, the specific instantaneous signaling of each element is independent of the others. It is only with the ordered regimentation of neural coalitions, those volatile and fleeting spreading activation patterns which result from the phasic extraction of contrast between connected elements, that the active properties of neural maps can be expressed.
The contrast between two instantaneous exteroceptive signals is an ephemeral, transient thing, and because it has temporal as well as phasic components, it cannot be expressed purely in the passive connectivity of neural or MAA ensembles. The topographic relation between two adjoining exteroceptive signals in a neural map is indeed expressed through their synaptic connection in the mosaic of the neural map ensemble, but the evanescent character in the contrast between these topographically related elements of instantaneous signaling can only be expressed as a neural coalition, a state of activation within the mosaic of topographic interconnectedness.
To be sure, if the process of contrast extraction was all that was necessary for a neural map to “compute” different centers of attraction in the whole of the exteroceptive signaling, then Nature would certainly have implemented connected neural ensembles to express this functionality of contrast extraction. But to graduate to an ability to discern separate shadows, to establish a technique which can abstract the constraints behind the veils of apparent nonlinearity, Nature will have to evolve a follow-on operation to the contrast extraction phase, in order to accomplish the subsequent stage of actual boundary resolution, a follow-on operation of “stitching together” the ephemeral patterns of individual contrast states, much like a meteorologist draws isobaric lines of equal pressure to create maps of weather fronts.
The bottom-up engineer might be surprised to find that Nature will not have to perform too much experimentation in order to evolve the dual stages of contrast extraction and boundary resolution needed to implement neural maps. Just as Nature adapted the metabotropic neurophysiology of persistence to implement first and second order state synthesis for abstracting the emotive character of photonic sensation, that first order state synthesis faculty would be ready made to implement the temporal and phasic processes of contrast extraction once the primordial neural map structure had evolved, a map which was itself a natural extension of the compound eye matrix, and certainly the second order state derivatives of metabotropic persistence used for the refined emotive schemes would perfectly fit the requirements needed for the second stage boundary resolution operation of connecting the individual, transient contrast coalitions that form the fluid boundary pattern.
Once the boundary resolution process had evolved, the metabotropic neurophysiology would again be brought to bear in the final step of segregating individual shadow “fields” within the photonic neural map by their boundary separations, a step which will mark the first emergent production in our evolving primordial metazoan.
The dialog will provide more detail on how neural maps produce the emergent behavior of global boundary resolution in local instantaneous signaling because it should be apparent how the survival value of the neural map will greatly elevate the probability of the recurrence of this new neural structure in many of Natures’ subsequent designs.
While the bottom-up engineer can envision how the adaptations which first evolved this primordial neural map would subsequently allow our metazoan to develop survival behaviors in those environments that presented multiple shadows, students of MAA’s should also see how the emergence of the neural map will have other, far-reaching phylogenic influences on the evolution of our primordial metazoan, as it evolves from the Cambrian metazoan into the Devonian vertebrate.
GO TO NEXT DISCUSSION
GO TO TOP OF DISCUSSION
Copyright © 2019 All rights reserved