Consciousness, Meaning & Memory
In order to conceptualise the brain as a goal-oriented machine, there must at least be goals that exist at some supervisory level. The goals at the third, uniquely human, cognolinguistic level are semantic states. Humans and animals (who do not, strictly speaking, have a third level) all use semantic states as the principal data structure in their brains (see memory map in figure 4.1 below). These semantic states exist in the form of percept categories- basically, data types. It is no accident that the data types used in computer programs are called their 'semantics'. Because of common coding, both perceptual and conceptual kinds of brain activity are stored as semantic states (perceptual categories).
The technical term for meaning is semantics. An important part of GOLEM theory is its reliance on a stricter foundation of linguistics. In GOLEM theory, linguistics existed in the mind first, then was externalised as language. Semantics is formally defined by an equivalent Moore machine (a synchronous type of finite automaton). Syntax is also formally defined in terms of an automaton, but as Mealy machines.
The relationship between the semantics and syntax of both external (linguistic) and internal (cognitive) types of information is one that is completely and unambiguously determined by these automata. Syntax is the cause and mechanism of semantic state transition. Each GOLEM Moore [1] machine consists of linked semantic states whose transitions, when they occur, do so simultaneously. This is by design, since semantics is a combinatoric code, with each state represented by a different combination of symbols. For this reason, Moore machines are called 'synchronous'. Each semantic state consists of multiple microstates. For the purposes of analysis, the micro-transitions that link these individual microstates must occur in some concrete order. Each set of micro-states and micro-transitions is represented by its own (syntactic) Mealy machine. Unlike the Moore machine, the transitions are accompanied by (symbolic) input and output. This is an unusual application for Moore/Mealy machines, but otherwise there is nothing mysterious going on. It is the same decomposition that occurs when a (synchronous) combination of symbols is expressed as a series of constituent (sequential) permutations of those same symbols.
There are some (non-language) real-world processes which correspond to these finite automata. Their detailed description assists in understanding the actions of the automata involved. Some i-semantic (ie internal memory) states each correspond to a somatic posture, or array of limb positions w.r.t. the torso. Note that (a) the limb positions are expressed as joint angles, which makes good evolutionary sense, since angles are an inherently scale-independent metric, and (b) these angles are divided up into a finite number of non-overlapping ‘buckets’, eg bucket<1> = {0 ->15 degrees}, bucket-2 = {15-30 degrees} [5] and so on. If a somatic state (ie a body’s configuration) is completely defined by 60 joint angles, then there will be up to 60 microstates that must change for each macro-state transition [6].
Each voluntary movement consists of two steps, the first one voluntary (Moore machine -master) and the second involuntary (Mealy machines-slaves).
(i) construct a 3D path (trajectory) by constructing a series of 3D points in space in the cerebrum.
(ii) execute the animation in the cerebellum, using the paths points as ‘keyframes‘.
Figure 5.1
syntax encodes semantic change not semantic state
In GOLEM theory, syntax is also defined by an equivalent finite automaton, but this time it is an asynchronous 'machine' (a Read Only Memory, or ROM) called a Mealy machine. This leads to much confusion, since the linguistic definitions used in GOLEM theory differ (slightly) from those in the scientific mainstream. Syntax is normally thought of as the surface form of a message's meaning, or semantics. Noam Chomsky is famous for his 'syntax-first' attempts to find out how language works. In his ‘minimalist program’, Chomsky seems to suggest that semantics is the 'deep code' behind language’s 'surface' syntactic form [4].
In GOLEM theory, the relationship between semantics and syntax is determined precisely by their finite automaton formulations. From comparison between semantic's synchronous Moore machine and syntax's asynchronous Mealy machine, we can see that syntax doesn't encode absolute semantic state, but semantic state change. When I speak to you, or write these words, I am not introducing completely new meanings into your tabula rasa mind, I am updating the mental image you already possess. This revision of linguistics clears up several nagging issues such as the difficulty of creating a Montagovian [2] (first order logical) interpretation which works with anaphoric forms.
Moreover, this formulation of i-syntax and i-semantics (the internal, cognitive versions of the familiar externalised concepts) leads to straightforward behaviour governance mechanics. By pursuing certain semantic states, which are already represented (or are representable) as percept categories in cortical memory, the semantic differential between current and desired states can be found. This leads directly to generation of syntactic formulae, since syntax is (re)defined as semantic state changes. In serial form, i-syntax can form spoken or written language, while in its parallel form, the multiple i-syntax streams can be used for concurrent control of the bodies effectors (its torso, limbs and musculature).
consciousness is evolution's gift to animals
There is a key insight about consciousness that arises when it is compared to meaning- consciousness is level-2 (ie spatialized) meaning, and conversely, meaning (eg of speech, or text..) is level-3 (ie symbolized) consciousness (see dashed red circle in figure below). First came consciousness, of course, which was evolution's gift to animals. We are conscious because we are animals, we inherit it from the animal world. Evolution's gift to humans is of course language, which can be conceptualised as an externalised form of thought.
language is evolution's gift to humans
This comparison between meaning and consciousness is not superficial- they are constructed in the same way, using a similar type of data structure. Each semantic state is pluripotent, which is why semantics is represented in memory as percept categories, eg as sets of exemplars. This property arises directly from the way semantic state encodes reality, as combinations of constituent properties [7]. This is why there is no easy conversion between syntax and semantics- syntax, as traditionally conceived, describes symbol permutations, a sequential code, while semantics is a combinatorial one. The semantics of any given situation is given by the combination of the semantics of its components. Semantics depends on the selection of semantic constituents present, not their order or sequence.
This correct view of semantics reveals how words work. Words are recursively semantic, each word represents a semantic category- the word 'table', when it appears by itself means ALL tables. Only when it appears in a sentence with other words is its semantic range curtailed. In fact, all the words in any given sentence exert a mutually restrictive effect on each others 'pragmatics' (ie in situ semantics). The effect is not permutational, but combinational. If you imagine each word as its own little Venn diagram, each containing all the exemplars that conform to its denotational (dictionary) definition, then the total sentence semantics is given by the overlap set. Clearly many less (semantic state) exemplars will remain after the combinatoric function has been applied. This view of linguistics has a name - the theory of underdetermined semantics [3].
consciousness incorporates (justified true) belief
But we have seen that consciousness and semantics are formally similar, in that consciousness is a spatiotemporal (level-2) version of semantics, a cognolinguistic (level-3) representational sub-type. By the argument above, semantics is seen to be pluripotent, including all possible exemplary states, so by extension, so is consciousness. This means that when we are conscious of the ‘current’ true state of our surroundings, we are also (subconsciously) cogniscent of all the other possible ‘next’ or ‘other’ states, those that could have formed our present reality, if things had been slightly different. This, then is how the mechanism of belief operates, according to the theory of Justified True Belief (JTB).
(a) (b)
Figure 5.2
In figure 5.2(a) above, a modified version of the 3-level GOLEM map is used to compare meaning and consciousness. Figure 4.2(b) depicts the TDE (TDE Differential Engine), which is the cyberanatomic analogue of the GOLEM .
Like the GOLEM, it has three levels of representation, but unlike the GOLEM, it contains fractal branching which mimic the brain's neuroarchitectonics. Note the duplex data paths to the language (Broca's and Wernicke's) Areas on the left side of the TDE, but also note how there are matching duplex data paths to the prosody/empathy Areas on its right side. The TDE details are a close match to actual brain electrophysiology (EEG, nMRI etc) as well as the observations which arise from localised trauma (lesions). This model successfully predicts that right cerebellar lesions cause Broca's-like aphasia [8]. This is an unusual prediction to make, suggesting that the model is substantially correct.
1. named after Gordon Moore, Intel's founder
2. after famous linguist Richard Montague
3. Vicente, A. & Martinez-Manrique, F. (2018) Semantic Underdetermination and the Cognitive Uses of Language
4. Notwithstanding his herculean prosecution of the 'minimalist program', Chomsky has been singularly unsuccessful in the pursuit of its closure, a fact that inevitably casts doubt on his findings. Nonetheless, his notions of external and internal merge correspond closely to changes to GOLEM's semantic hierarchy.
5. the number of 15 degrees has been chosen arbitrarily, for convenience of explanation
6. if these angles are to be changed by parallel neural processes, they must be relative 3D angles, not absolute 3D angles, because the latter type DO NOT COMMUTE. Because relative angles (eg roll, pitch,yaw) commute, the order in which they are changed is unimportant, meaning they can be parallelised (timesliced) without changing the overall computational result.
7. According to Pulvermueller (2013), “Referential, combinatorial, emotional-affective, and abstract semantics are complementary mechanisms, each necessary for processing meaning in mind and brain”.
8. Silveri MC, Leggio MG, Molinari M. The cerebellum contributes to linguistic production: a case of agrammatic speech following a right cerebellar lesion. Neurology. 1994;44:2047-50.