There are hundreds of billions of neurons in the human brain and body, but it has been shown that only 80 billion neurons are associated with memory and higher reasoning.
It will be shown that your brain would not have enough neurons to store your memory, if memory were stored chemically.
It has been proven that it takes 70 milliseconds for a neuron to fire, and reset itself to an electronic potential that will allow it to fire again. 70 milliseconds is 70/1000th of a second, so it is possible for a neuron to fire about 14 times each second.
So, if we assume that the brain is a multiphasic neural network, rather than a single value neural network, where the weights and purposes of the individual nodes can change from firing to firing, rather than having a set value, then 7 times as much information can be stored on each node. It would be 14 times as much, but in a multiphasic neural network there have to be selection commands, so each node knows which new value to change to for the next firing sequence, and the selection command firings have to be at least a 1 to 1 ratio to information processing firings.
So, this increases the available storage space to 80 billion neurons times 7, or 560 billion neurons.
Your eyes have 200 million rod and cone structures each. It has been shown that the brain only processes information from 10 percent of these at any given moment.
The speed of television, 30 frames per second, is the minimum frame or sampling rate, because slower than that, the eyes are not fooled into beleiving they are witnessing a continuous moving picture.
So, out of a total of 400 million rods and cones, the basic continuous resolution is 40 million times 30 samples per second, or 1.2 billion individual bits of optical information per second that the eyes present to the brain.
Say you are 35 years old. By the age of 35, your brain has processed 1.2 billion times 60 seconds times 60 minutes times 24 hours times 365 days times 35 years worth of information, divided by 2 to take sleeping and blinking into account. This is 1.324512e+18, or
662,256,000,000,000,000 bits of information.
560,000,000,000 (560 billion).
There is a million times as much information to be stored as there is available storage space.
If one assumes that visual memories are stored in only a tenth the resolution in which they were originally perceived, there is still 100,000 times as much information to be stored as there is available storage space.
Remember your last dream? Remember what the blackboard looked like in first grade, what the eraser looked like, what was printed on the 26 cards posted around the upper part of the walls?
You be the judge. What resolution do you presume is the minimum resolution that visual memory can be stored in? Given the quality of your photographic memory, the smaller and further you get from a 10 percent figure, the more absurd it is to consider... Any less than that and you wouldn't be able to differentiate one person's face from another's.
It has been shown that the brain is a neural network, composed of large numbers of neurons which serve as the nodes. These nodes take input values and produce output values as a result. Each can fire as many as 14 times per second.
With 80 billion of them, 80 billion individual bits of information can be being processed at any given time. So, the brain has the capability to process 1.2 billion bits of visual information per second, but not the capability to store it chemically. In a neural network, each node has only one function. It receives voltage and possibly pulse, pulse width, and timing information, and produces an output pulse as a result. In a multiphasic neural network, more values can be analyzed and processed, in the case of the brain, as much as 14 times as much information.
However, it is more likely that the brain is a standard, single value processing neural network, rather than multiphasic, because with a multiphasic neural network, there have to be command pulses or a command structure implemented in the weights of the nodes. Many operations have been performed in which a part of the patient's brain was removed, in some cases as much as half of the brain, in patients with severe epilepsy, with no discernable memory loss whatsoever.
While there might be duplicity inherent in a standard neural network, it cannot be implemented in a multiphasic neural network in such a way that would prevent memory loss, or the command structure's duplicity overhead, both electrical and physical, would occupy so much of the available resources that in a case where there is only a 14x improvement in storage capacity by having multiphasic capability, the vast majority of the 14 percent improvement would be occupied by the command structures, and one would be left with a network that could contain far closer to 2 percent as much information than 14 percent.
If your memory is stored at 10 percent of the original resolution, then last month, you processed
6,570,000,000,000 bits of visual information
1,120,000,000,000 Even if the brain could store 14 times as much as 80 billion bits, you can see that it still isn't enough to store more than about 5 days worth of visual information.
Only in a multiphasic neural network can a node (a neuron) have variable processing capabilities. If one were ignorant, one might say that the memories could be stored chemically inside the neurons by the billions, billions of bits of storage inside each neuron, but it is seen that this cannot be.
A neuron receives input, processes it instantaneously, and produces a voltage pulse as output. There are no calculations involved. There are neurotransmitter chemicals that provide communication between neurons. When enough of these chemicals have been received by a neuron, the neuron fires.
It has been seen that when enough of these chemical have been received from all the neuron's inputs, a point is reached where the voltage potential becomes high enough to fire the neuron. The neuron does not have the capability to differentiate it's inputs. If a neuron has 25 dendrites (inputs), and it receives a total of 60 millivolts of electronic potential from these 25 dendrites, it will fire. It does not matter which of the 25 inputs provided the input voltage. If the total potential presented by any combination of the 25 reaches 60 millivolts, the neuron will fire. This has been demonstrated through experiment, as have any facts stated on this page. Your basic human physiology textbook can be used to verify these facts.
The voltatge potentials have been shown to flow along the outer membranes of the neurons, and not through the neurons. Without the flow of current travelling through the neuron, there can be little influence of the potentials by matter present inside the neurons.
There is also the problem of the index. Neural networks are associative. Each memory in a true neural network is connected to and associated with similar memories. Each new memory takes all past memories associated with the given subject into consideration, is assigned a new weight, and then stored. The problem arises from the fact that if we are 25 years old, and attempt to remember what the blackboard in grade school looked like, we have to step back through 20 year's worth of memories to find the location where the memory is stored.
Since we can only perform, at most,
1,120,000,000,000 associative recall operations per second, and we have to scan back through
1,576,800,000,000,000 bits of information to find the location where the memory is stored.
or, in other words, it would take 1000 seconds to locate the memory.
Of course, there might be a timebase reference in operation, that gets used as a basic index to avoid the need to step back through the linear chain of data association.
However, although it would make much more sense to use a timebase, reducing the access time to a microscopic fraction of a second, it would require a dedicated physical architecture, and people have had half of their brain removed without memory loss. The memories themselves, if stored chemically, would have to be themselves spread out over a wide area of neurons in the brain, but the timebase would also have to be. The only possible physical layout of a timebase index is a three dimensional chemical vector table matirx, with physical connections to the storage vectors and the actual storage locations. It would be possible to remove half of one's brain, but not remove the center of the timebase structure, but there would be considerable damage to it's outer physical connections, and there would be large and significant gaps in the memory structure, resulting in years' worth of memories being lost. No such memory loss has been witnessed. Also, since memory is associative, all later memories, not some later memories, those memories which were created after those in the damaged area, would be greatly modified. If the memories were stored chemically, they would be stored as combinations of neuron weights, as all other values in any neural network are. Once the chain suffers damage, the value of all the connected affected nodes change.
One can say that a great deal of reduncancy and error protection can be built into a neural network, and this is true, especially true for computer based neural networks. This is not as true for the neural network found in the brain, simply because there aren't enough neurons present for it to be implemented to any significant degree.
Another angle is the processing overhead. This is the most obvious and undeniable of all arguments against memories being stored on billions of chemical bits inside neurons.
Think about it. A neuron can only fire 14 times per second. The optic nerve delivers 1.2 billion bits of information per second to the neural network.
In a neural network with substantial error correction and redundancy, the data is stored redundantly in its entirety in a number of different associative nodes. But in the case of the brain, only 14 bits of information can be stored on a neuron each second, and there are 120 million bits of memory to be stored if visual memory resolution is 10 percent of the original.
So, the information would have to be spread around, and stored on at least 8571429 neurons each second. This itself doesn't prove that the memory cannot be stored on chemical bits inside the neurons, only that each visual memory would have to occupy a minimum of 8571429 neurons.
However, 8571429 neurons is a considerable amount of widespread neuron real estate. It is only 1/9334 of the total number of neurons available, but if you go back to the minimum required timebase layout, it becomes obvious that no two memories can have the same neuron as their starting point, or the storage location for their first bit. The timebase must exist, because if it didn't it would take you 1000 seconds to recall a 20 year old memory.
The optic nerve is continously delivering information to be stored. The timebase and the optic nerve cannot both access the same neuron at the same time, the timebase to read or write and the optic nerve to write. This is a physical impossibility, because only one operation can be performed by one neuron firing. Therefore, the timebase must by the nature of the physical and electrical layout, assign sequential neurons as starting points for the memory that is to be stored there.
There must also exist two seperate electrical pathways to the neurons, one for the optic nerve, and one for the timebase structure, because it has been shown that exposure to electomagnetic fields increases cell membrane permeability The small electromagnetic field generated by the neuron firing from the optic nerve would interfere with the values coming from the timebase index as it records it's vector values. Cell membrane permeability is the basis by which adjacent nerves pass chemical messages.
We have other types of memories. We know what a hot dog smells like, what it feels like, and what it tastes like. Most of us have higher reasoning facilities, otherwise known as knowledge and wisdom.
We had to learn how to walk and speak, read and write, etc... Higher reasoning bitwise memory storage would make the storage space requirements of visual memory pale in comparison.
Copyright 1998, Robert J. Nelson.
See also COPYRGHT.GIF, an integral component of this work, for a full legal copyright notice.
This work is composed of the following copyrighted files: COPYRGHT.GIF, NUKE.HTM, DNA.HTM, and BASICS.HTM. No other files are covered under this copyright.
You must have the author's express written permission to reproduce this work if either of the following are true:
1) It is part of any package sold for profit, or if you gain monetary profit from the use or publication of any part of this work.
2) It is your intention to republish this work, or if you publish this work, in any medium which contains advertising of any kind.