In memory of Joe Egan, born 25th February 1916
Here we move tentatively into the terra incognito of the physiology of value perception. Expect amendments to follow.
The previous posting on Ubiquitous Error Elimination leads us to consider that value perceptions gained through the automatic numerical meanderings of a Least Squares Method fitted model may have something in common with value perceived by real consumers of some actual goods.
A point of commonality is to consider one’s brain as a model that delivers consciousness of observed reality. The observations are sensory inputs that comprise many analog signals, which an internal perception must seek to organise into a cognitive model with a minimum error value. Although the physiological processes are largely unknown, one might expect some form of neural modelling and fitting to observed reality as part of the emergence of conscious awareness. If so, then all such model fitting will be subject to the general principles of navigating a complex error surface.
In The Grand Design (Bantam Books, New York, 2010) Stephen Hawking and Leonard Mlodinow set out to address some very big questions employing “model-dependent realism”, which assumes our brains form models of the world based on information received through the senses. There is no definitive true reality and many such models can co-exist and may be adopted dependent on their usefulness and value. Such individual perceptions of an external reality are likely to depend on intrinsic assumptions and will probably find acceptable limits of error that are sufficient to ensure survival. To do more, at least in a primitive society, would be a waste of energy.
If one could link the concept of a potential field that we have hypothesised as acting to constrain the creation of value in the raising of a Value Surface, to an error surface associated with cognition and perception, then this could indicate some physiological foundations for the earlier hypothesising. Some physiological potential must be driving a descent down an individual’s cognitive error surface to achieve a reliable perception of reality. Otherwise one’s consciousness would have no physical cost and Maxwell’s Demon might happily defy the 2nd law of thermodynamics. Information that confers survival, in a primitive society, and perhaps quality of life in a modern society, must be classified as more valuable, indicating a higher use-value to external objects or a greater exchange value, than a random replication of useless information. Hence information can be made valuable through the very processes of biological perception.
This sketching out of an association between the physics of value perception and the biological origins of consciousness is entirely speculative. A deeper analysis of this association must await another day. However, the general concept of an intrinsic neural model that is fitted to observed reality does lead to some interesting observations.
This concept explains individual differences of opinion that are reflected in an oscillating Value Surface. Individual perceptions of the real world clearly differ. The start points to fit a neural model to these perceptions certainly should differ. Different acceptable local minima on the error surface may provide different individuals with a different interpretation of the same reality. Human beings probably do not process sensory data exactly in the same way and clearly can reach different conclusions when given similar scenarios to manage. If people behaved as automaton robots, then each commodity Value Surface would be a rigid plane of equal valuation.
Yet people’s opinions and beliefs are extremely stable for such a dynamic fitting of internal model with external reality. Such stability could arise if the start of every new minima search begins at the most recent minima for a comparable reality, perhaps retrieved from memory. In this case only the change from the previous reality needs to be reconstructed in the modified internal model, which then provides the next start point in a continual modification of an internal neural model to reflect changing perceptions of a real world.
Whilst cogitating on this very subject of the fundamental origins of the conscious mind, the author was sitting on a bench on Barcelona waterfront. A very brief interruption was made by a smart thirty-something year old who mixed languages rapidly in an urgent attempt to communicate. Within ten seconds the chap had disappeared along with a bag containing everything that was valuable, snatched by a second person during the distraction. Passport, wallet, travel tickets, laptop, money all had disappeared. Yet in disbelief I imagined I could see my familiar grey stolen rucksack where it should have been, on the bench beside me, for a good few seconds before grim reality fully allowed itself to be recognised. Reality had changed too quickly and it seemed the refitting of my internal model was taking long enough to notice the processing delay. Later on at the UK Passport Office, I was informed that Barcelona is the bag-snatch capital of Europe and had I known this, the adjustment to the new reality might have been smoother. Or maybe I would have protected my belongings with more conscious deliberation.
Several years on, the memory of that minor Barcelona trauma is fresh and easy to recall.
As considered in “Writing the Information”, can such vivid memories be the river valleys etched into the error surface of my consciousness by the cascading experience of these earlier events? This is a subjective and even metaphysical suggestion, but such a cognitive system should certainly be an attribute in favour of survival and as such could be a selected epigenetic trait. Important information would be considered valuable by its hosts. I will be more careful of my luggage on any future visit to Barcelona.
Whatever are the mental mechanism and however current controversies on the nature of the mind and consciousness play out in the future, the subject is central to understanding innovation. Not only are the intellectual processes that act on information at the very origins of innovation, but the subjective appreciation of value by the consumer, whatever the product of the imagination, can be traced back to its source in the obscure processing of the human brain and its constituent 100 billion information processing neuron cells.
Back in Barcelona in 1887 Santiago Ramón y Cajal started to work with a new Golgi staining method that used a silver preparation which, for the first time, enabled neurons to be clearly visible through a microscope. It was the start of the modern discipline of neuroscience. Ramón y Cajal used the Golgi method to produce many graphical illustrations of complex neuronal shapes. On observing these cellular structures exemplified below, it is difficult not to see similarities to the dendritic patterns considered in “Writing the Information”, and to infer that the associated metaphor might extend into this neuroscience domain. That is, the tree-like neuronal patterns once again infer, albeit circumstantially, that an energy transmission function is at the heart of these microscopic constituent cellular elements of the brain and central nervous system.
Santiago Ramón y Cajal shared a Nobel Prize with Camillo Golgi recognising their work on the structure of the nervous system which today forms a “Neuron Doctrine” that is a basis of the current understanding of the anatomy and physiology of the central nervous system.
Drawing of Purkinje neuron by Santiago Ramón y Cajal, 1899;
The dendritic structure of neuron anatomy and physiology enables the cellular behaviour to be mapped onto the generic “Green Box of Innovation” template introduced earlier. In this case, an electrical signal flows from the multiply connected and complex dendritic structures, through to a central cell nucleus and a single axon strand of connected cells that can reach across millimetres, to stretch out to a branched terminal region, there to connect to dendrites from neighbouring neurons. The axon-dendrite connection is known as a synapse in which a communicated signal is transferred by chemical means. Here the information transfer through the synapse requires a transformation of electrical to chemical energy in neurotransmitters and then back to electrical energy as the neurotransmitter binds to synaptic cell receptors to begin the transmission through the next dendritic link of a connected neuron.
A synaptic link connecting neurons can either excite or inhibit the transmission of an electrical signal, known as an action potential in connected dendrite links. Perhaps 10,000 such dendrite signals converge on a cell nucleus to give rise to a single event which occurs at the axon hillock, the point where the filamentous axon connects with the cell nucleus. This integration of the many dendrite signals that need to cross an energy threshold to determine whether that neuron will fire a single electrical pulse through its axon to communicate with its cellular neighbours is a main physiological function of the brain and other parts of the central nervous system. These pulses may last for only a millisecond and each neuron may contribute to the information flow up to 100 times per second. Clearly there is much information flowing through the average brain.
We have described an energy transfer process that needs to reach a critical threshold before a neuron will fire and propagate its signal. Billions of such signals must converge to create a perception of value at a Consumer Product Interaction that is a precursor of a Sale Event. Again this is an integration of received information into an “all-or-nothing” decision to purchase. Though differing in terms of scale similarities appear in the energy flows of the action potentials of neural circuitry and those operating on consumer preferences in the shopping centre.
There are Artificial Intelligence (AI) models that attempt to replicate on a very small scale the manner the brain naturally might function. Neural networks, an example of which is shown in the figure below, are brain-like numerical models of layers of connected neurons whose connection properties provide a generic set of parameters that can be specified to characterise the behaviour of the system. These connection parameter values can be estimated by using a Least Square Method, navigating to the lowest point on an error surface between a simulated behaviour and a known “training set” of real output values. Once the ideal simulation with the smallest error has been found, then the associated neural network parameter values should faithfully reproduce the real world, so long as this is retained within the limited confines defined and exemplified by the training set.
A typical neural network connecting four input neurons to two output neurons
through a single intermediate layer of 6 intermediate neurons.
AI neural networks can be useful as they continuously learn from new data just as humans might. The predictions they can make can be informative as are human intuitive predictions. They are also susceptible to weaknesses of ambiguity in human understanding. There may be many local minima on the error surface to trap the descending Least Squares Method. Also, like the brain, a neural network model is adaptable to fit with the many diverse challenges an organism might face, but this means the solution is an arbitrary fit to observable data. There is nothing intrinsic in the model that represents the world that is being simulated, nor are there any overt assumptions that can intelligently be applied to simplify this real world.
In the real brain of the analyst, the real multi-billion neural network can be applied to explore the world using models with some conceptual simplification. Effectively this is positioning the human processing power at the front end of the entire modelling process.
This is the origin of An Innovative Enterprise Simulation that uses the Method of Least Squares to provide a vision that would otherwise be unavailable to the unassisted human senses.
It is a model to explore the process of innovation itself.