Brain Physiology, Cognition and Consciousness group: topic
This is a public discussion board
From Neuron to Memory System: How Memory Might Work
Graeme Smith
Friday, 12 June 2009 04:34 UTC
How does memory work?
We think we know, based on thousands of years of research into it, but there are still indications that we might be completely wrong. Scientists created computers with the best scientific approach to memory based on current scientific theories, in the 1940’s, and they got it wrong. Most neuroscientists today will tell you that the computer is nowhere close to a good model of memory, yet many of the attitudes that informed the decisions on how to make computer memories remain unchallenged. Our popular theories of how memory works were laid down in the middle ages, by people who thought memory was a fluid, and that the brain was a reservoir. Somehow they thought, we opened valves, and let the memory flow from the brain to where it was needed. I call this a demand model of memory because you demand the memory by opening the valves. In computing terms we have replaced the fluid theory of electricity but retained the idea of current, and at one time tubes which were replaced with transistors were called Valves. So computer memories could be seen to be operating in a demand memory model as well.
As Jerry A. Fodor said in The Mind Doesn’t Work That Way!:The scope and limitations of Computational Psychology Neither Phenomenal approaches such as Neural Networks nor strict computational approaches such as Truth Preserving Functions, seem likely to achieve a suitable computational model of consciousness.
As a researcher in Artificial Consciousness, my main thrust is to eventually get support for my own Artificial Consciousness to be built. However to get there, I had to start with a model of memory. I chose to hedge my bets, to begin with a Neural Model, and add Functional support to that model, where it was needed. Further I wasn’t going to limit myself to Truth Preserving Functions, where Soft computing would be more effective. This type of approach to consciousness is called a Hybrid approach. To start off, however I wanted a model of memory, and one of the things I had to do, to get there was overcome Fodors assurance that no-one had yet developed a phenomenal version of a demand memory.
Those who are stuck on William James idea that phenomenal systems can’t be built based on what we know about the brain, have problems with my use of the word phenomenal in the way that Jerry Fodor used it in his book. They like to look for exotic things like QM entanglement to explain the fact that some things just seem indivisible. I however think that there are indivisible elements in the memory but that it doesn’t matter because the mind doesn’t try to divide them, it finesses the system, without that effect.
Fodor’s claim is not without merit, phenomenal systems are definitely easier to build out of neural networks, than demand memories. However, David Marr defined a type of memory he thought might help explain the cerebral neocortex back in 1970, and by understanding what his work uncovered, and brushing off some of the unfortunate assumptions of his day, I think I have undercovered one of the hidden biases that is keeping us from understanding the way the brain, or at least the memory works. And that bias is the assumption that explicit (demand Model) memory, is the natural state of the memory system, and that anything that doesn’t fit the model must be an add on to the basic explicit memory model. So we expect Implicit Memory to somehow be an add-on to the essential demand memory, and we are wrong.
What Marr described in 1970 is a type of memory more basic than explicit memory when implemented in neurons. To computer guys like me, this seems counterintuitive. How can anything be more basic than a good demand memory, we can implement dynamic ram with a capacitor, a resistor, and a transistor, how much more basic can you get? But what we keep forgetting is that neurons don’t work like transistors, there is no technological overlap. Within the logic that makes neural networks work, demand memory is much more expensive to build than a simple content addressable memory. That is what Marr claimed his 4 layer CODON was, was a self-classifying content addressable memory.
To understand why this might be, we need a little theory. Although in the 80’s and 90’s the connectionist school was over-run by the Parallel Distributed Processing guys from the A.I. discipline, and so we have to take some of its theory as being deliberately misleading to steer people away from trying to understand real neural systems, The basics are fairly equivalent. A neuron is first of all a cell, it has to survive like other cells, by absorbing nutrients and getting rid of wastes. However at some point in the evolution of animals, neurons gave up some of their survival functions to a helper cell called a glial cell, and converted those functions over into mechanisms for transferring information between cells.
The Parallel Distributed Processing guys figured what was important to this communication was the firing of the cell. Sorry that is a misconception, firing just speeds the process of transferring information up, it is not the only mode of transfer, nor is it the most important one for understanding natural systems based on nerve cells. However because the PDP boys wrote the manuals for the industry, they got to tell 20 years of modelers what to think. As a result of mistakes like this, they set the neural modeling of Natural Neural Networks back, to the point where Gary Marcus, has clearly stated that he feels it necessary to do a hostile takeover of the connectionist school.
One approach that the PDP guys thought was dangerous was David Marr’s attempt to use probability Mathematics to define a circuit that was made up of heterogeneous neurons. Despite the fact that they did not have the ability to model heterogeneous groups of neurons, a direct attack on his claiming that the heterogeneous group he called a codon was a content addressable memory, was made pointing out that the model did not exibit wave-forms similar to the real cortex.
This attack, assumed that Marr’s circuit had to have frequency artifacts similar to the real cortex in order to suggest the role of content addressable memory for the cerebral neocortex. Well Marr was a pioneer in the field and he got some things wrong, and others even more wrong, but it was a seminal stage in the science, and what he got right is more important.
Marr was able to use probability mathematics to analyze heterogeneous networks of neurons and predict their function. Nobody before that even tried, and after the way he was treated few after him had the nerve to try again.
I am not going to start spouting probability equations, if only because I don’t follow his math. But the PDP guys called a neuron a processor that had input, process and output capabilities. I however look at it from a different perspective I see a neuron that has storage, processing, and transfer capabilities. And when I look at the 4 layer cortex model that Marr worked with, I also see a content addressable memory, but I think that perhaps he expected too much equivalency between neurons in the connections department, and was sadly dissapointed before his death in 1980. So I wonder at how much self-classification the system is doing.
If you look at neurons from my viewpoint you begin to see that the shape of the neuron has a logic, that memory neurons have lots of synapses, and opportunistic process growth, that processing neurons tend to have complex and often bushy or mossy dendrites, and that transport neurons tend to have virtually no division at the dendrite level and form long thin neurons.
If you look at the type of organs that are produced, Memory Neurons tend to form tissues, Processing Neurons tend to form globular organs, and transport neurons tend to form fibrous bundles. In other words the shape of the neuron, indicates it’s function, and the shape of the organ might indicate to some small extent what type of function it performs.
Looking at Marr’s 4 layer model, we see that the first layer is processing neurons, the second and third layer are memory neurons, and the fourth layer is dominated by processing neurons again. The whole structure is probably a memory role because it is a tissue.
Stimuli coming in at the first layer, are processed affecting their storage in the second and third layers, and the 4th layer probably is mostly control neurons, and in some cases inputs of senses that have been directed through the thalamus. These latter signals condition the general inhibitive environment caused by the Martinotti cells and thus reduce the inhibition at the first layer encouraging the layer 2/3 neurons to fire.
This simple memory circuit can be seen to be designed to respond to either signals from layer 4 or layer 1, and the first layer is used to train the second and third layers to respond to patterns of stimulus. Given a stimulus each pyramidal cell in layer 2/3 decides whether or not to fire, and this type of system is exactly what is needed for a content addressable memory.
It is also exactly wrong for an explicit memory suggesting that it must be the explicit memory that is the add on. Indeed when we look at the micro-architecture we see that a type of tissue called Allocortical tissue makes up the tissues at the bottom of the Sulcys and Divides of the brain. and since Sulcys and Divides are a later architectural configuration there might be an indication that the six layer tissue called isocortical tissue, that pushes the gyruses up away from the sulcys and divides, was a specialization of the brain.
-
Replies
Jump to resultsResults
-
I’ll class that in the “You can’t get there from here” vein of thought.
I think the poster made some good points, although their comment was so scattered as to be almost unintelligeable. I especially liked the way they included motorboat, in a discussion on memory;)
This model so far, has been looking at the memory from the point of view of the types of cells that make it up, and conveniently ignoring the connections that link the neurons. My only treatment has been some comments about the bushyness of the dendritic trees for a particular neuron type.
The poster is doubtless correct to doubt that computers could completely model the complexity of the connections involved. The cerebral cortex has billions of neurons, and each has literally tens of thousands of connections at the synapse level. Obviously your average PC is not going to be able to do a single pass through a whole brain model, in any reasonable amount of time. Estimates are that this might change in about 20-30 years if computer manufactores can keep up the frenetic pace of development for that long. If the pace picks up, which it might, they might shorten that period, but for now, consider 20 years a good estimate for being able to simulate a mouse brain. Considering development requirements that means the theory that we are going to use to develop the software has to be produced some time in the next few years. I am hoping to be a part of that development.
At issue in the development of a whole brain model, is one factor that I have mentioned before, the indeterminacy of the address of a specific memory within neural networks. This occurs for a couple of reasons. 1. the opportunistic nature of connections between neurons, 2. the Phenomenal Nature of Neural Networks as suggested by Jerry A. Fodor in his book The Mind Doesn’t Work That Way!: The Scope and Limitations of Computational Psychology.
The idea that we might map the brain has existed for centuries, and has been proven wrong time and again. This seems counter intuitive, if we all see the same color red, and have nearly similar DNA, then why don’t we have similar maps of our brain? The answer as Dr. Edelman mentioned in The Remembered Present: A Biological Theory of Consciousness is that DNA does not fully specify neural shape. It meerly provides a framework in which neural shapes can adjust to fit the environment. Gary Marcus recently noted something similar in his writings which include the book Kluge which notes that DNA sets the rules by which growth can occur, but doesn’t directly control the growth outside those rules. The brain is almost chaotic in its organization at the neural connection level. Yet it overcomes the chaos by the time it reaches the Gyrus and Sulcys level of organization because brain surgeons have shown a facility at mapping the brain at that level, that is not shown at smaller scales.
So if we can’t map a memory to a neuron, how does the brain find it’s memories? The first thing we have to accept is that the brain doesn’t store a memory in a single neuron. Instead, there is evidence at least in the cerebral cortex, of structures in the cortex that promote the grouping of neurons into clusters of about 100 or so neurons that act together to promote one or two neurons to fire. This means that despite chaotic connections, at the neuron level, groups of neurons can become experts on a particular aspect of a memory, and can vote for which neurons fire. Because we know that the memory circuits involved voluntarily fire, we can say that the memory is not the output of a particular cell as it might be in a computer, but the output of all the cells in the data cloud that respond to a specific stimulus. Thus at the implicit memory level, the data-cloud is the phenomenal memory, and the individual neural group, is just one aspect, one viewpoint by one part of the brain. If we think about this we see that addressing at the neural group level, won’t give us particular memories, just bits and pieces that might suggest larger memories, made up of many such bits and pieces, to the brain.
-
Dear Graeme,
I admire your enthusiasm but am cautious about yet another person saying that they can build a model of a brain like never before. Having been in Cambridge neurophysiology when Marr was at Trinity in the late 1960s I can see reasons for going back to him. He may still be 10 years ahead, even if nearly 30 years gone. And I agree that he left unfinished business, and perhaps blazed some false trials. I wondered about getting the Fodor book you quote but the view I have formed of Fodor’s approach so far is that, like Putnam, he has got some very basic things muddled up and I am not as yet convinced that this book will disabuse me, even if it introduces me to some new use of terms.The difficulty all of us on the list may have is that in your enthusiasm you throw out a lot of terms expecting us to know which of several meanings, and ontological categories, you are giving them: memory, phenomenal information, data cloud, voluntary firing. (You also have a penchant for shattering illusions about the brain that most of us have, I suspect, never held, but no matter. Many of us on this list are fairly hard boiled, having delved into these crannies before, to mix metaphors.) You seem to be homing in on an issue that I would consider of central importance – that certain cells are involved in ‘explicit’ or ‘phenomenal’ (experiential in my terms) brain events where others are either not, being mere number crunchers or housekeepers or perhaps handling data in some sort of ‘inverse’ Fourier-like arrangement. You seem to be trying to link the phenomenal to heterogeneous cell function and structure, which is precisely where I think we need to go. Yet I have yet to see the (non-quantum) leap beyond Marr and your usage of terms like phenomenal or voluntary in the context of ‘data cloud’ and your comments about such clouds being indivisible worries me that the Putnam mirage might lurk beneath. Please disabuse me by giving us a concise definition of your versions of the words in the first sentence of this paragraph. With that we could get a dialogue. It may not be an easy task but without it there is no great incentive to continue on the thread.
The other word that has raised its head in the thread is ‘representation’, and if you have a meaning for that it would be interesting. Where I thought Marcus fell short in Algebraic Mind was in steering clear of what representation actually means. There seems to be an assumption in AI that everyone knows what it means so there is no need to say. Yet the Kosslyn/Pylyshyn debate, while only scratching the surface, shows that there is no such consensus. How did ‘perceptrons’ come to be called that and what was the coiner thinking at the time about their relation to ‘percepts’, or did that get brushed under the carpet?
There is a can of worms in here. You may agree about that and we may be able to debate on the same wavelength. On the other hand you may agree but we may remain unable to engage. But if you define your terms we will know where we stand. My terms are defined in published papers quoted on my webpage (search for Jonathan Edwards at UCL).
Somewhat surprisingly, yet not for the first time, I think Otmar’s last little message hits the nail on the head. Maybe one day Otmar and I will find we are on the same wavelength after all.
-
OK, so maybe, Graeme, if you had pointed us to http://en.wikiversity.org/wiki/User:Graeme_E.Smith/Dual_ModeMemory
everything would have been crystal clear from the start (did you?). In the Wikipage you define your terms and as a result are easy to follow. The non-quantum leap is clear in the sense of the dual mode memory. It rings very true. It even rings bells with something I have been trying to figure out on paper for a while with limited success. I call this (for my own benefit) a mordant loop which ‘fixes’ a circuit that can be entered at two points, one subserving the question ‘can you remember Anne?’ and the other the question ‘whose face is this?’ It requires at least four types of cell in at least three brain regions. I never read Marr on cortex. I know his cerebellar model and the vision book but I have focused on more general issues since being interested in the brain again (thirty plus years on). Your use of quale is seriously confusing because you use it as a dynamic term and the usual use has nothing to do with dynamics. The way you use it is fine in context but it is likely to confuse most people in the consciousness field. My own view is that the other non-dynamic meaning of qualia does help to lead one to look in the right place for the dynamics but if you have a handle on that anyway then forget it for now. Put it another way, I suspect that your model might end up pointing to the same dynamic architecture as I would want to see to solve William James’s very real logical problem about binding. There might be a little surprise involved, but that’s for another day.Thanks for the bibliography on the Wikipage which I will enjoy exploring.
-
Good responses guys.
First of all, I have been remiss, partly because I am a little shy about bandying my actual papers about. But for those of you that didn’t catch it in Jonathons post, I have a page at Wikiversity Home Page where you can read a few actual articles I have written that might explicate this idea better than I am doing on this thread. Some of them are under review at the moment and may change, but it will give you a bit of a background.
Ottmar, the longer we talk the better your questions seem.
As for Representational Realism I don’t claim to be of that school of thought.
It’s complicated by the nature of the way that we store data about the world. Yes we store real stimuli, but we do so by reinforcing existing views of similar stimuli. So what we are storing, is not exactly what we sensed.
It is a mixture of what we are familiar with, and what we make sense of, by reinforcing existing memories of past stimuli. This effect slightly warps the nature of memory, making it easier to remember familiar things, and harder to remember novelty, at least at the implicit memory level. Since I am still talking about the implicit level and its interface to explicit memory, I won’t speak to the other forms of memory that do deal well with novelty, at this time. It is valuable because it makes it possible for us to classify things in a similar manner to how we classified them earlier when we first stored them, but it is hardly realism.
RR in general but not specifically RR. Let me explain Karmiloff-Smith developed a system having 4 memory representations, I have already suggested that I have read an article that denotes 5 representations. What I actually believe in is a system with 6 or more representations. Considering that I believe that we have at least 9 attention systems even this is probably too small a set of representations to fit the complete model.
Not all representations however require redescription. Some of them merely require remapping or symbolic links. A six representation system based only on those representations that require redescription, is however plausible even in a 9 phase attention system. To be redescribed I say that the memory has to require it’s own new memory area, and a new representation that cannot be achieved directly by mapping.
Ok, now on to knowledge. I haven’t got anywhere near there yet, but I will try to explain it when the time comes. I want to stay strictly away from the thought level however, because I am not trying to analyze what thoughts are, I am trying to synthesize a memory system. Thoughts come after consciousness in my plan. First I want to make a consciousness about as smart as a rat or rabbit, then I will worry about teaching it how to think. Suffice it to say, that like many things, processing often defies intuition, and designing a thinking machine might be much easier if we can get the fundamentals right for the cognitive architecture.
Innateness is one of those things that Fodor and Karmiloff-Smith disagree on to some extent, they both think in terms of modular memories, but Fodor assumed innate knowledge, and Karmiloff-Smith was able to show how some of the memory that Fodor would have thought innate, was actually developed by the brain as a response to the stimuli, it received. I won’t go into the experiments with animals that were raised with a patch over one eye, except to state that there were distinct differences in the development of their brains in comparison to animals raised with both eyes open. In other words there was information from the environment that helped determine the hard wiring of the brain, and thus what Fodor might have mistaken for innate knowledge.
Karmiloff-Smith does recognize innate knowledge partly I think because of her background with Piaget, but she maintains that it is much less prevalent than Fodor expected.
Jonathon actually I don’t recommend Fodors book, even though I cite it. I found one gem in it, but in general it was just another set of lectures by a self-important modularist.
I will get into my own interpretation of Modularity later, when I explore deeper into the Attention System. But I think that when I get there you will find that it helps with the question of why some cells are memory some are processing and some do some pretty exotic functions.
As for defining representation, how could I possibly be so bold? From a computing mind-set, when we represent something, we code it so that certain characteristics of it are easily available to the program that we hope will use it. For instance we don’t store real numbers in a computer we store codes, that have a symbolic meaning, that can represent numbers. Part of the theory of how we do that is based on Information Theory, which was developed by the phone company. When we code a digit in base 10 sometimes also called decimal format, we have several possible coding standards from which to pick.
The A.I. Guys, are always trying to develop computer “Representations” of knowledge we know the brain processes, with the idea that if they can find the right way to code the data, and can define the right manipulation system, they can do the same things the brain does, but with a faster system.
People like Dr. Edelman are strongly against this viewpoint, and even Fodor, has expressed some reservations about it at some points.However even Dr. Edelman and Fodor seem to agree, that there is a difference between simulation of a natural system and direct coding of knowledge. Both have mentioned that neural networks create a phenomenal system, because the encoding of the information in neural networks is not based on information theory, but on relationships between neurons, and that is a much more slippery concept. Whether it is Dr. Edelman’s claim that consciousness is a phenomenal gift, or Fodor’s claim that neural networks have not defined a demand memory yet because they are phenomenal systems, and do not lend themselves to selection of discrete memories, let alone location of those memories at a specific address, The implication, is that neural networks are more phenomenal than functional.
There is a trend in Cognitive Architectures towards mixing the ideas of the Phenomenal school like Edelman, and the Functional School like Baars, and that trend has been labelled Hybrid. I am firmly of the opinion that the hybrid approach is superior to both Phenomenal and Functional approaches at building a consciousness. However Dr. Sun who designed the CLARION Architecture, and I do not see eye to eye about how we should design such an architecture. His approach is much more oriented towards Functional theory, while my approach tends to favor Phenomenal Theory. I would build a phenomenal memory system and only thne add on functionalism while his idea is to put a few layers of neural networks to implement implicit memory and to do the rest with functional modules.
I think there are some places where we are going to get into cans of worms, and I don’t necesarily want to go there if I can help it, which is why I was so short with Ottmar a time or two.
-
Dear Graeme,
Would you mind explaining why your version of memory or any version of memory, for that matter, has phenomenal properties. Quite frankly, this seems to me like an end run around a whole lot of functionality.
Michael Baggot
-
The phenomenal worm-can, at least in Edelman-Fodor form seems to be pretty hard to keep shut. Michael’s indigestion is likely to be shared. I would personally like to unpick this because I think it may be a useful lever. If people like Edelman and Fodor are using phenomenal to indicate a type of dynamics then I would like to know their reasons, since at present nobody knows what dynamics equate to phenomenality. My suspicion is that there is a false premise, but I am always keen to discover that false premises are not so false, or disguise something more interesting. I also sense, as indicated before, that they are discussing a real problem about the dynamics that a brain needs to work the way it does and although phenomenality is not directly relevant it is extremely relevant in an indirect way.
The comment I would focus on is “neural networks create a phenomenal system, because the encoding of the information in neural networks is not based on information theory, but on relationships between neurons”. (I am unclear whether this refers to biological networks of brain neurons or the connectionist models that are called neural networks. I will assume it is intended to relate to real brain neurons.) What ‘relationships between neurons’ are being invoked? As far as I can see the only relationships between neurons that can be relevant to the causal processes of interest are relationships of information passing – the local relations of each neuron to all those synapses providing it with input in a dendritic tree for instance. The only other relation between neurons that I am aware of is there position in space, which is of no causal significance (you could move them about without affecting brain function one iota) so cannot encode anything we are interested in. The only difference in relations in brains as opposed to computers that I can see is that in brains neurons can relate to thousands of input synapses whereas computer gates only relate to two inputs.
So my question would be, does either Edelman or Fodor specifiy what encoding relationships they are talking about?
I am also curious about the idea of mixing phenomenalist and functionalist accounts because as far as I know these are incompatible ontological positions rather than mix and match patterns of dynamics. It is a bit like the Mexicans mixing Catholicism and Maya culture. It can be great fun but does not seem to have any coherent intellectual basis! If phenomenalists are represented by Edelman and fucntionalists by the AI people I strongly suspect both are equally following beliefs that on close inspection fall to pieces (but am happy to discover otherwise). A hybrid between two incoherent frameworks would be a sorry mess. A hybrid of two sorts of dynamics, on the other hand, is perfectly fine, if there is a dynamic justification, and I would agree that there probably is.
-
I think I see the problem now, I am mixing ontologies. Phenomenalist Consciousness Ontology, and Phenomenalist Philosophical Ontology. They are both part of what I think is a larger ontology, that is amenable to synthesization with physicalist/Neural Network means, so I didn’t bother to separate them in my mind. I am not sure they need to be separated, but if you guys maintain that they are distinctive enough in your minds, that they deserve to be separated, then since I am trying for a consensus here, I will bow to your wisdom.
It is entirely plausible that I was misinformed into mixing the Ontologies by Fodorian error, since as a philosopher he should not have mixed the Consciousness Ontology into his Philosophical Book, but that is history.
I still maintain that there is a larger ontology of which each of these is just a symptom, and that simularities between the symptoms merely show that there is a commonality that links the two systems, but that is open for debate. I suppose I could have been fooled into linking them through the name of the processes involved.
However, Jonathon, I am not sure the static/dynamic distinction is the right one to make between these two ontologies. Why do you think Philsophical Phenomenalism is static? For the purposes of this discussion just to have a label to stick them with, I will call the original ontology Philosophical Phenomanalism, and call the new Ontology Network Phenomenalism, is this acceptable to you?
So, yes, I suppose unpicking Network Phenomenalism might indicate whether or not there is a fallacy hiding away here, or some real insight, that is just awkwardly expressed, not the least by me. After all I am the poor slob who is trying to use that argument to base his theoretical model on.
To answer the questions you have raised, I think I need to go back to hebbian Neural Models, where a Synapse was seen as an expression of the strength of the connection between two neurons. So not only do we have lots of equivalent synapses, but each synapse has a value called in Neural Networks a weight that defines the strength of the relationship at that particular connection. (It’s why it continues to be called Connectionism even though we now know that we can’t map the actual connections.)
Now you may wonder, whether I am talking about neural networks or real neurons. In Neural Networks, we model real neurons, so the question is one of how close to the natural neuron is our model? The answer is pretty far, but getting closer. However if we interpret the model, back into biochemical reactions, we can get an idea of what is really happening within the real neuron, even if it might not be completely accurate.
It turns out that for every permease molecule in a synapse there are a number of ion channels that pump ions in or out of the cell. The number of ion channels, pumping depends partially on the number of permease molecules that have been activated, and partially on the ratio of ion channels to permease molecules. We can see that this ratio, is in fact, just what is needed to define a value we can call the weight of the synapse. To carry the model forward the number of ion channels in the sensitive patch of a neuron, is adjusted by synaptic activity and its relation to other biochemical factors in the neuron. One of these factors is the denaturing of the proteins that make up the ion channels by their eventual digestion in the cellular fluid.At some point enough of the proteins are digested that the ion channel no longer functions. To keep this from happening too soon, the cell has a maintenance mechanism that replaces the cellular membrane from time to time, and this mechanism has the ability to sub in a bright shiny new ion channel from storage, to replace an old and half-digested one.
If this replacement mechanism is triggered early, it adds an extra ion channel, and every once in a while it removes a defunct ion channel. The important thing is that the rules that determine which happens are triggered by the activity of the synapse. While the actual chemical circuitry is not available to me, because I am not a bio-chemist, what is important is that the weight of the synapse here represented by the ratio of ion channels to permease molecules in the sensitive patch, is adjusted as the cell learns to reinforce active synapses.
thus what our memory actually stores is information about the relationships between neurons especially those that are local and fire at about the same time.
It is because the neuron is not storing data about the signal, but data about the relatinoships between the signals, that I say that it is storing information about the relationships between the neurons. Perhaps I should have qualified that and said it is storing information about the relationships between the Activity Levels of neurons, would that have been more accurate?
Either way, whether we are talking about the Neural Network model of the Neuron or the Biochemical Model of the Neuron, we are still trying to model the very complex nature of Natural neurons, using ontologies that are not completely incompatible, and the same concept seems to fit both Ontologies.
I suppose we could say that this implies that the Neural Networks, at least in this case are a superposition of the biochemical Model. But I may be using the term too loosely.
-
Micheal, I don’t know why you think phenomenal memory systems are an end run around functionality, if anything they are an indirect approach, and require more functionality to perform. What I am hoping to accomplish is to make an end run around our lack of understanding of phenomenal memory systems, and thus to make a system that performs the tasks despite a lack of clear models. The reason for going with a Hybrid system is that we understand some things well enough to functionally model them, but others less well. The simulation approach is really just a stop gap while we are figuring out the functional model. I merely suggest that we know a little less than current functionality thinks it knows, and so we need more simulation rather than less. Thus my flavor of Hybrid Memory System is based more on simulation than on modeling.
-
I would like to remind everybody that this is an international forum. You guys are mostly native English speakers and people like me are not. Moreover there is a problem of philosophical, neuroscientist etc. jargon.
The discussion is great, but could you please try to communicate in a more simple generally understandable way. (this is not meant to suffocate your creativity, just a request)
-
Just one remark about my understanding of the relation between phenomenality and memory.
Phenomenal is what we experience. Things, experiences appear to us in certain ways. – I could start an aside here about how Alfredo and I have proposed to reconceptualize qualia (in our JCS article) here, but I don’t do it. -
After the moment of experience, some part of it or something like a part of it may enter memory. That is one part. The other part is the process of remembering. In that process a new phenomenal experience is created.
It may in some cases be very much like the first experience, but in many cases it has got to be different. This is how I see the nature of human phenomenal consciousness and memory.
Results
-
