Long-Term Memory, Simulated Qubits, Physical Qubits - Brain Theory from a Circuits and Systems Perspective - page 57

Chapter 4
Long-Term Memory, Simulated Qubits,
Physical Qubits
Introduction
The five senses transmit a large number of signals to the brain, and after they are
processed, they present an image in conscious short-term memory that is composed
of fundamental attributes, that is, specific colors, shades, shapes, edges, tones, and
so on. The existence of an particular attribute in a given image may be indicated by
a simple true (present) or false (absent). Each image in long-term memory is thus a
binary record. Additionally, each attribute of each image in long-term memory will
have to have links to conscious short-term memory, so that recall is possible.
Memory, be it short and long term, depends on bistable memory elements, one
for each attribute, each holding true or false. This is in contrast to other fanciful
concepts. For instance the grandmother neuron is supposed to hold an image of
one's grandmother, we are told. But it is difficult to imagine how a grandmother
neuron would be searched among the maze of other such neurons, and then
transferred into conscious short-term memory. To avoid such practical problems,
this topic stays with the concept of one neuron per attribute in what has been termed
distributed memory [ 1 ].
More is known today than say, a decade ago, but much remains unknown about
the physical details of human memory. Reasons are, memory neurons have minimal
impact on their surroundings. Neurons are efficient with energy compared to solid-
state devices, and so taken individually, they do not release very much heat and
radiation, making them difficult to trace. They are microscopic; they are tightly
integrated with billions of other neurons, and so are difficult to isolate in action.
In vivo experimentation on a connected neuron is especially challenging since they
are easily rendered inoperative by probing.
Considering possibilities for long-term memory elements, there are differing
purposes. A basic element stores a signal for the presence of a given attribute,
and returns its information with 100 % fidelity, a simple true or false. Recursive
neurons, on the other hand, store a basic logic value, but can return their logic value