Brains of Sand N Neural Schematics
Figure N.1
Figure N.1 above depicts the net flow of information in the vertebrate (including human) brain from perceptual input to motor output. The diagram does NOT depict common perceptual coding, ie the translation of all motor codes into their sensor-side equivalents, even though this constitutes one of the major features of the TDE model. The diagram also does NOT depict the crossways flow of information between the input and output channel, an information flow which is known to be large and important. These omissions are a deliberate strategy at this stage to simplify the explanation. Later we will see how the cross-channel links which enable the transverse flow of information form the foundations of phenomenal (ie subjective, or self-governed) agent spaces at the 'hardware' (ie 'wetware') level .
The purpose of the diagram is to demonstrate clearly, directly and in an absolutely simple manner that a complex brain can be constructed from circuits containing only one type of neuronal element* - a standard multiple-input-single-output common-or-garden neuron. The fan-in (convergence) required to create the 'bottom-up' input representation hierarchy, and the fan-out (divergence) required to create the 'top-down' output reproduction hierarchy are both easily achieved with the standard many-to-one neuron design. Note also, that even when a neuron's axon is connected to multiple targets , such as the parallel fibers(PF) in the cerebellum- each PF's's axon intersects many purkinje cell dendrites- the output signal at each of the axonal targets is identical**.
The basic information processing pattern of the single neuron is up-scalable, ie it is a fractal building block of the brain's neural circuitry. In other words, a single neuron has the same circuit I/O morphology (though not the same connection topology, of course) as does the compound, multi-presynaptic-single-postsynaptic unit made from three or more single neurons (see thumbnail graphic located in the lower-right part of figure N.1).
*In a roughly similar way, all modern computer hardware circuit logic is reducible to circuits made only from NAND (Not-AND) gates.
**like a single radio with a speaker installed in each room in the house, with volume (but not station tuning) separately adjustable for each room's occupant.
What are the benefits of using this conception of neural data flow? As with most of TDE theory, they are summarised by the ORPH3US^ principle. Perhaps the most important advantage is that it allows a physical interpretation of the linguistic basis of cognition* (LBC). LBC is based upon the idea that all brains are semantic state machines with syntactical (syntagmatic?) state transitions. LBC permits that brains can be (in theory) modelled as Turing Equivalent Machines (TEMs) of some type. LBC also leads to an interpretation of the declarative computing paradigm as multi-level goal-oriented processing (ie drive-state reduction). This is as simple and direct as it is bioplausible^^. It is pleasing to be able to present such complete definitions of syntax and semantics after almost two centuries after Humboldt**^^. They feature (i) denotational and connotational compliance/closure (ii) confident cross-topical use eg for synthetic cognition, theoretical computer science^^^^.
Consider the input channel, which uses neurons connected in a 'feedforward' manner** to create an ascending perceptual, or representational hierarchy. To make this hierarchy a representation***, we need it to possess the following attributes (a) intentionality (b) nested or recursive constituency. This is most easily achieved by (i) equating hierarchical level with physical scale, and (ii) letting the children of the parent node at that level be its feature names, implemented as categorical (content type) addresses in content addressable memory (iii) features are composed of superliminal attribute combinations^^^. In linguistic terms, we only need the input hierarchy to be equivalent to nouns or noun phrases (adjectivally qualified or compounded noun groups). The input hierarchy performs an integrative function, not dissimilar to a taxonomy. Each child node has a unique antecedent, a single parent category. There is no multiple inheritance at this stage of the model. This reflects the essential (ie unique, non-ambiguous) nature of intentional semantics.
Now consider the output channel, which also uses feedforward neuronal interconnections, but this time, the multiple connections are all (individually scaled, if required) copies of the neuron's single axonal output. Each time we descend a level of the conceptual, or reproductive hierarchy, we must make choices as to which children (parent's subcategories) to tokenise (ie make copies of, if the model is discrete, or to choose the scaling parameter, in the analog case). These transitions correspond to changes of semantic state, which are linguistically equivalent to syntactical expressions**^^. Like syntax, they imply choice, ie formulaic selection of figure/signal and rejection of ground/noise. Chomsky laid the groundwork for this type of operation, which involves algebraic combinations of categories, called grammatical formulae or 'production rules'. Each rule interprets each nth level*^ macro-semantic symbol (called a 'non-terminal' and written in lower case) in terms of its (n-1)th level constitutive symbols or semantic atoms (each called a 'terminal' and written in upper case). Because the element of choice is mandated, syntactic operations are all voluntary. That is, they require an agent (or source of agency, typically a subjective entity**^ or self) to generate and pursue goals, a process called conation or drive.
^ the hybrid modelling heuristic/aesthetic consisting of Occam's Razor (OR is the simplest framework which accounts for function) and Pierce's Hook (PH is a.k.a. retroduction, and involves choosing the axiomaton ie inherited set of precursory assumptions or beliefs about the system, which offers the most 'unremarkable' explanation of its observations/behavior). ORPH3US is a 'hackronym'- Occam's Razor Pierce's Hook Hybrid Heuristic Uncommon Sense. The reason for using the hybridised form of the heuristic/ aesthetic is to acknowledge the significant overlap between the results of applying OR and PH, probably due to the fractal nature of all information processing mechanisms, irrespective of provenance (ie both biological and artificial).
* perhaps the most extreme form of LBC is that claim that thoughts=words. However, many mind theorists do seem to promote a less severe version of this identity, attributed mainly to Jerry Fodor and (to a lesser degree) to Noam Chomsky, namely that (a) thoughts, like speech, are macrosymbolic, (b) they derive combinatorial macrosemantics directly from their constituent semantic atoms (thoughtlets?).
^^=biologically plausible.
**ie whose branches don't have closed loops with feedback flows
*^These recursive relationships which act like stairways, connecting adjacent floors of a linguistic high-rise, if you will.
***defined here as a data structure with parametric similarity to a perceptual image or 'reality window'.
^^^Treisman, A. M., & Gelade, G. (1980). A feature-integration theory of attention. Cognitive Psychology, 12(1)
**^Note the deliberate avoidance of the dangerously ambiguous word 'conscious' in this context. Proper use of the term 'conscious' describes superliminal (above threshold level) entities or actions in the representational hierarchy ONLY. The usual colloquial use of 'conscious' is not incorrect, but risky, because of the latent suggestion of agency, or volition. When Benjamin Libet decided to use electrophysiology to investigate the phenomenon of 'conscious' choice, he overlooked this semantic 'stowaway'. This leads directly to an 'impossible' interpretation of the data, apparent evidence of the reversal of time itself. Re-examination of the data using ORPH3US reveals the flaw to be semantic- the improper use of 'conscious'.
**^^Humboldt. On Language, On the Diversity of Human Language Construction and its Influence on the Mental Development of the Human Species. Edited by Michael Losonsky, CUP 1999, pp. 25-64
^^^^wider adoption of these definition awaits critical discussion & acceptance from academic community heavyweights.