On learning

Learning is an “unconscious, contunuous automatic process” within the brain. It just happens when the conditions are just right.

The most fundantal factor is repetition, which implies by doing, which is the current meme for restrictive ancient practices.

A practice is a millennia-old concept, and some advanced cultures consider every activty in life (and life as a whole) as a contunuous practice of [spiritual] virtue.

Again, repetition is the “how”. There are, however, a few “optimization hacks” about what to repeat and when and how often.

The biological basis of learning is so-called neuroplasticity - actual changes within individual sinapses and axons. There are molecular “markers” which trigget the actual stuctural changes.

The neuroscience meme is “fire together – wire together” sums it up “not wrong”. The same “principles” apply to a muscle tissue. The “brain is a muscle” meme is biologically incorrect, but conceptually “right”.

So, just as in literally any sensory-motor skills acuisition, be it playing a musical instrumet, dancing or athletic performances, everything boils down to “what to repeat and when”.

The details of the processes of changes within synapses and individual neurons are well-understood by modern neuro-science - no contradiction with the ancient principles.

It seems that the generalized abstract notion of a converging spiral-shaped (recursive) process captures it “perfectly”.

Well, individual synapses or neurons are “not recursive”, but the continuous (repeated) refinements are indeed “converge” to a “local optimum” and “platoe” with “diminishing returns”.

Again, sensory-motor skills (even in aninals) have its limits, and “convergence” (of abilities) slows down and eventually stalls.

All these concepts are proper generalizatins and related obviously to each other. Even artificial neural networs (which is just “a curve fitting”) demontrate the same dynamics.

To write down is to clarify ones thinking. To write down generalized mathematics (to use a precise language) is an optimum.

Most of people, including myself, write down and then forget, because we do not revisit our wirtings and do not clarify (re-write) what we wrote.

This is a fundamental mistake, just like stopping exersising or switching to another exercise way too early. Jumping between activities to early and too often is a definitive symptom of ADHD.

So, writing notes (or even flash-cards) is absolutely required (for neuro-maturation) but it is not enough. Neglecting the repetition (of re-reading and clarifying) is the cause of descuraging, poor performance.

This is what we do with “neural circuits” inside our brains, our captured and properly generalized concepts, our texts, and our programs. And this must become a habitual practice.

Just like running or shooting arrows.

The informal definition of The Right Understanding (of the Buddha) is the absense of any contradictions to What Is - to an already partially-solved “Jigsaw puzzle”.

All non-bullshit generalized concepts can be traced back to What Is (to the frequently observed phenomena they has been geralized from). The ability to zoom back and forth between the concepts and the actual aspects of the shared environment is what distinguish an intelligent mind.

This is one of the most ancient principles, which goes back to at least Upanishads and early Buddhism. Being “grounded into what is real” is a universal “safety net” and THE most fundamental principle of all.

It captures the very same notion as training of modern neural networks - your training dataset must be absolutely “real” and well-choosen, and exclude all kinds of “glitches” and “garbage”, otherwise the result will be a garbage.

Notice how universal these notions are - all higher living forms “train and maintain” an inner “map” of the shared environment (“territory) insede their brains and use these ”maps“ for decision making (instead of the environment itself).

This is an operational definition of an intelligence.

Unlike neural nets, however, human minds are not merely do “statistical learning” (from repeated experiences), but experience is still the basis of everything (along with pre-defined, evolved specialized brain areas).

So, watch your “inputs” and guard an already trained “neural structures” as the most precious thing.

I used to be a heavy FreeBSD user and abuser. Nowadays I am a Gentoo guy (which acutally tells a lot). Both systems are based on compiling individual components (packages) from the source code, and both have “continuous” updates.

Some would argue that this is a wasteful practice (an if measured only in a time consumed, this is, indeed, the case), but as a “side effect” one develops an intuitive undestanding and such falimiarity with the system, that there is literally no fear or even uncertainty about what is going on.

The Right Understandig is what required when dealing routinely with the dependency hell, and this very understanding (of what is actually going on in the whole open-source ecosystem) is what we are after.

The point is that just doing repeated task develops abilities beyond jsut “muscle memeory”, it also alters the doer.

Bring in information from multiple sources: multiple textbooks, Wikipedia, video lectures and written lecture notes (from Univercity websites), etc.

The more subtle sources include vocabulary, slang and the common idioms from the language of the professional in the field – the up-to-date vocabulary of the problem domain.

The ultimate goal is of “extracting” the right set of properly generlized concepts and actual relations among them, which, again, do not have any contradiction to what is already “known”.

This is the ultimate hacking - the knowledge extraction (sort of a reverse-engineering from the mountains of verbalized bullshit) and subsequent use of what one has distilled and purified.

Adapted from this blogpost https://borretti.me/article/effective-spaced-repetition