Discussion 10: The Second Fundamental Precept

Discussion 10a: The Root of Self-Organization

From its inception, it has been the goal of the Organon Sutra to develop an engineered plan to instill intelligent behavior in an artificial agent. An initial first step in this engineering can be found in a classic dilemma in computer science, which asks “How can a programmed computer reorganize itself in response to its inputs?”

A programmed system does not have the capability to be selective about its inputs, because by definition, it will process those inputs it has been programmed for in the manner in which it has been programmed to process.

In contrast, complex biological systems that are exposed to the infinite variety of the real world do not have the capability to process the entirety of their simultaneous sensation, which is transduced by a sophisticated array of systems with phylogenetic sensitivity, and therefore these biological systems must be selective about their inputs.

The ability, or actually the necessity, of complex biological organisms to selectively attend to varying subsets of their inputs based on current environmental salience and not on phylogenetic programming lies at the root of the self-organization that programmed computer systems are incapable of. The foundations for this conceptualization in the Organon Sutra was hinted at in discussion 9d, when the dialog introduced the sensory function of perceptual diffraction, but the mechanisms for that functionality have yet to be detailed.

By design, each subsequent state of a programmed digital computer system is defined for any number of time steps following the start of execution of its program, and any allusion to self-organization within the machinations of a pre-programmed system is merely a sophist deception, however the allusion is couched in so-called artificial intelligence design. It is because of this that much of the artificial intelligence community has embraced connectionist models, although this dialog has detailed many of the limitations of that modeling approach.

Self-organization within the realm of contemporary digital logic systems can only come about if the system has the ability to alter its own programming based on the exigencies of its real-time inputs, but how can engineers design a self-programming systematic that does not statistically follow the entropy of its environment and degenerate into digital chaos?

To answer this question, the Organon Sutra turned to the singular example of natural intelligence, the human central nervous system for clues, and in discussion 7, the dialog specified that all epistemic activities of the CNS stem from the biological adaptation performed which, at its most fundamental level, is a process of moving entropy from the organism out to its environment. It is the maintenance of this fundamental process that must form the essential goal direction in a self-programming systematic.

Therefore, the characterization of this “entropy”, and the maintenance of the entropic process within self-organizing systems became central to the definition of artificial intelligence as the design of the Organon Sutra progressed.

And in the previous discussion, the dialog continued an imagination scenario depicting the early evolution of motile metazoan species in their development of a locomotive axis, and their requisite development of neural mechanisms that collect instant information from the environment in order to steer the motive axis accordingly.

To this point of the imagination scenario, both the evolved contact electro-chemical sensory modality and the non-contact photonic sensory modality had developed neural structures having the sophistication to derive direction vectors towards their particular sensory idiom, the direction of gradient concentrations in the case of the former, and the relative direction of shadows in the latter. And we saw how photonic sensation evolved because of the limitations that contact exteroception placed on mobility. Finally, the previous discussion concluded with our imagination scenario at a crossroads, created by the inability of photonic sensation to signal any aversive or attractive character in the idiom it was signaling. This crossroads was cast as an evolutionary selection pressure to develop neural mechanisms physiologically different from any of the neural elements so far evolved by Nature.

However, there was another enormously important aspect to this new neurophysiology as well. Just as the mechanisms that collect the instant information of the environment that our primordial metazoan was using to steer its locomotive axis increased its chances of obtaining energy or avoiding aversive conditions, it was paramount that this collection machinery did not undergo any functional modification in the process of its sensation. The reaction to environmental signals from all sensory modalities had to be accomplished in a repeatable fashion, irrespective of any signaling activity that occurred prior to it.

This biological imperative represents the principle divergence in neural conceptualizations between the Organon Sutra and traditional Hebbian doctrine, a doctrine that mistakenly treats every synapse as equally malleable to the very signals that it transmits, and a divergence which introduces the conversation to the Second Fundamental Precept, and the neurophysiological considerations of state in the neural processes of organisms.

Researchers can only speculate about the degree of physiological functionality in neural cells as they had evolved for organisms in the pre-Cambrian and Cambrian eras. Based on Natures’ predilection for electro-chemical exteroceptors, the assertion that chemical transmission via neurotransmitters at the synaptic cleft having been well evolved by this time is beyond question, although the variety of neurotransmitters expressed is certainly open to debate. But even at this early stage in neural cell evolution, Nature did not have too many knobs to tweak in order to implement a new neurophysiological method to express neural circuits that did more than just transmit the integration or differentiation of their incoming impulses, creating a new phenomenon which would lead to the elementary abstraction of state in the activity of organic sensation, and it’s very significant derivative, the retention of state in neural activity.

Because so much early research was focused on isolating the “neural correlates” of memory, or the retention of state, it is not difficult to see how so much of neuroscientific research overlooked the formalization of state in neural signaling to begin with. A neuroscientific understanding of the phenomena of memory has until now remained elusive, and the profusion of theories and models that attempt to objectify the phenomena gives testimony to the need for a prior formalization of state abstraction in the first place. Although all activity in a massively asynchronous assembly is in its principle form the signaling of thresholds, the fundamental first step in abstraction is moving from signaling to state. And this must be accomplished only while the organism is exposed to the environment it will be abstracting.

One of the reasons that the search for the neural correlates of memory has been so elusive is because of the inverted nature of the traditional conceptualization of neural processing. Researchers proceeding from the a-priori premise that neurons are processing “something” that had to be “remembered” were putting the cart before the horse. We must begin with a different thinking in bottom-up design and MAA engineering. Neural activity must begin with signaling without context, and only then do we conjecture a mechanism to attach “information” to signaling, at the point of perception, and not at the point of ontogeny. It is only with a process of this true memory, this conjectured mechanism, do we bind specific information (the abstracted context of signaling) to a discrete computing device. This clarified process of memory ends with a neural state that traditional researchers begin with, and as the details of the Third Fundamental Precept unfolds in this dialog, we will see how this revised meme solves the dilemma of recall and “memory retrieval” that bedevils so many theories and models of “memory”. Starting out with a computing device that processes some “information” greatly complicates the process of “remembering” it in any fashion.

Although this conjectured mechanism seems to contradict the edict that was declared in discussion 9c, which specified that the signaling of neural cells contains no attached “information”, the Second Fundamental Precept is intended to precisely resolve this apparent contradiction, and bring into focus the misconceptualizations that have frustrated the search for a neural correlate of memory for so long. Axonic action potentials are indeed stateless, but the dendritic thresholds that trigger axonic action potentials can have components which express state by virtue of a variable response in the post-synaptic membrane to a stateless pre-synaptic signal. It is this manifold state response that the Hebbian doctrine cannot express, since the doctrine does not separate the stateless nature of axonic transmission from the manifold state properties of neural dendrites.

In discussion 9e, the dialog characterized the first step toward intelligence as the evolution of a neural array capable of the apprehension and exploitation of the simultaneous spatial differences in the same sensory idiom. And prior to that, the dialog expounded the necessity that all sensory idioms also carry an emotive sense of aversion or attraction, in order to release any phylogenetically evolved motor behaviors developed for that exploitation. With the newly developing photonic sensation, however, the neural array had developed the ability to apprehend the simultaneous spatial differences in the photonic sensory idiom of shadow direction, but with no emotive sense to enable its exploitation. Nature would have to learn to abstract a synthetic substitute for the emotive component, and the path that evolution chose, among others, was the apprehension of simultaneous temporal differences in the previously apprehended spatial differences.

But how could Nature bring temporal differences into simultaneity? The resolution of the dangling vector component in photonic sensation requires developing a neural mechanism that responds to changes in the direction or magnitude of the apprehended composite vector, a response that requires some analog of memory.

Since all of the neural mechanisms that are gathering the instant information about the environment, along with the circuits that discriminate the simultaneous spatial differences which develop the composite vector, must all function without modification by any prior sensation or processing, introducing temporal behaviors within their neurophysiology would wreck the whole works. This is the principle objection to the Hebbian doctrine. Nature needed an entirely different neurophysiology for this new temporal behavior. And because the tonic assemblies that Nature had been experimenting with so far required a continuous linear stimulus to maintain resonance, their functionality was also ill-suited for this particular temporal behavior. The first physiologically different adaptation (as opposed to the purely functional adaptations that evolution had pursued so far) that Nature would evolve was the need for increased persistence in the very nature of signaling itself.

As a formal matter, this nature of persistence is not the temporal summation of multiple impulses, but at its most fundamental level is conceptually the “stretching” of a single axonic impulse across the span of a variable number refractory periods of the neuron, with the relative number of periods that are spanned determining the degree of persistence.

For the students of MAA’s and bottom-up engineers, this “stretching” of a single impulse signal is the neurophysiological embodiment of “state” in the neuron. And so before we consider the adaptations that would be involved in the first implementation of state in the developing neural array as a whole, it would be instructive to hypothesize how Nature would go about evolving persistence in the neuron itself.

If Nature were to begin these evolutions by adaptations to the neural axon, altering the transmission characteristics of axonal action potential impulses would have required radical changes to the electro-physics of axonal impulse propagation. By the time of the Cambrian era, evolution must have already solved the problem of signal degradation along passive transmission lines. Passive electrotonic spread of charge potentials travel fairly quickly along the capacitive neural membrane, but the charges also dissipate and fade just as quickly. Surely by the Cambrian era, Nature had evolved its electrogenic method of active transmission. Using phylogenetically engineered mechanisms called voltage gated sodium channels and voltage gated potassium channels as the ionic equivalent of charge amplifiers, Nature learned to actively propagate action potentials along the axons of neural cells with no signal degradation. Although slower than electrotonic spread, there would have been little latitude for tweaking the temporal behavior of this method, due to an exquisite balance in a self-compensating charge reversal between the two channels.

And it is hard to conceive of how Nature could tweak the temporal response of the pre-synaptic bouton once the action potential reaches a synapse. Pre-synaptic boutons release measured quanta of neurotransmitter, called vesicles, in response to the arrival of an action potential, and although the post-synaptic latency effects from a single vesicle saturation represents the most basic form of memory expressed by all neural elements, changing the time course of this vesicle release or the vesicle quantum would conceivably involve alterations to many other finely balanced synaptic mechanisms, such as vesicle endocytosis, which is the process of re-uptake of neurotransmitter back into the bouton, a process even now not completely understood. And alterations in the metering of vesicle exocytosis would affect the phenomenon of synaptic fatigue, an effect which already marshals the long term behavior of synapses, and the very interaction of neurons themselves.

So, almost by process of elimination (there goes that teleological thinking again), tweaking the temporal response to action potential impulses would only be workable at the post-synaptic membrane of neuron cells. Now, in the conceived neurophysiology of Cambrian era organisms, the next higher form of memory expression above single vesicle saturation is the phenomenon of threshold saturation, where dendritic depolarization is maintained across the refractory period of a neuron. But threshold saturation is achieved only by the cumulative activity of many synapses and by definition could not be effected by a single synapse.

It is here, at the post-synaptic membrane, that we must look to for adaptations that would effect the phenomenon of persistence that has been alluded to, and just as importantly, the “conjectured mechanisms” that might lead to an expression of state abstraction, and ultimately, the retention of state in our evolving neural array.

It has been said that the intrinsic quantum of the brain is the axonic action potential, and conversely, the intrinsic quantum of the mind is the dendritic threshold. The Organon Sutra will emphasize that the student of massively asynchronous assemblies must approach the dynamics of neural activity from both perspectives. Although the previous few paragraphs have brought the focus of our examination toward the behaviors of dendritic thresholds, bottom-up engineers must not lose sight of the circular interdependency that arises from the basic assertion that it is dendritic thresholds that trigger action potentials, and axonic action potentials that produce dendritic thresholds. Since it is so easy to get our overall design conceptualizations caught up in this endless circularity, one of the formalisms of MAA’s is an explicit dichotomization between the push and pull of these two behaviors. But like every other concept we have approached so far, this formalization has depths and dimensions that require a complex understanding. And now this added phenomenon of persistence is complicating that understanding even more.

And the Organon Sutra cannot express these complexities and added complications until a complete picture of first, second and third order neural dynamics has been painted, which will not be complete until the third and last fundamental precept is fully detailed.

In the meantime, an innate understanding of the reflexive nature between the action potentials of neurons and their dendritic thresholds must be developed in the conceptualizations of bottom-up engineers, conceptualizations that can frame the context for a discussion on state and state retention in the organism. For this, those students of massively asynchronous assemblies must be able to visualize the very processes that constitute basic neural asynchrony between the smallest of neural aggregates.

To permit this, the dialog will suspend the imagination scenario of the evolution in our primordial metazoan neural array that has been playing out, and the bottom-up engineer will be asked to envision an interim scenario in which we visualize the mechanics of asynchrony at the most basic level.



Copyright © 2019 All rights reserved