Long-Term Memory Neural Circuits, Fast and Precise - Brain Theory from a Circuits and Systems Perspective - page 140

from past events to set the first toggle T1 to true and to begin the filling of a given
cluster of blank words.
The Memorization Enable serves to enable the NOT gates shown. If the blank
word above a given blank word is at rest, as they all are initially, the AND causes a
toggle in the subject blank word from false to true. This true is applied to the subject
Memorization Control Line, which is made to cause an image to be written into a
blank word. Since what is being written is still applied, there will be a Lout signal.
This serves to toggle T1 back to false and to toggle the next T2 to true. So the next
blank word is now ready to receive an image as soon as Memorization Enable goes
true again.
Except for the first blank word, the others depend on Lout to modify their
toggles. Consequently they are not all toggled at once, only one at a time. For all
except one, their state is false, or at rest, so this approach is efficient from the point
of view of minimizing neural pulsing.
Memorization Versus Learning
Simulated Qubits in Savant Memorization
Gifted savants sometimes possess photographic memory; their calendar skills require
memory; and memory may be needed for exceptional mental calculators. But what
does it mean to have great memory? Memory is useless unless there are cues for
recall. Memories of random information, for instance, are more likely to be recallable
if they are connected together somehow by mnemonics, a form of encoding.
What was originally thought to be amazing memorization by savants may in fact
be rapid learning. State machine learning is defined to be a type of subconscious
learning that uses hidden pointers to produce a next step based on previous steps.
Unfortunately, this type of learning requires rehearsal time. After suitable rehearsal,
interneurons become involved and eventually, there is synaptic growth. But savants
with memory skills can look over a long text and recite it almost immediately and,
we are told, they can recall it years later. Apparently savants require far less
rehearsal than average in order to learn sequences.
Learning a Long Sequence
State machine learning is thought to be responsible for all of the many procedures
humans employ unconsciously, including routine actions and recitations without
conscious involvement, or contemplating. The big advantage of state machine
learning is that mental procedures, which range from the mundane to the advanced,
do not have to undergo recall processing through conscious STM each time they
are used.