Most generally, learning is the process of acquiring new memory
. In sentient beings the acquisition of memory may occur in a variety of ways, for example, through the process of experience
. The process of learning seems to include behavior
as a component (i.e., motion, or activity).
. . . . . . .
In Biological Neural Networks
We are just beginning to understand some of the underlying biological mechanisms of learning. These include PTP
, a silent-synapse concept has also been offered.
Various interactions between proteins and other factors seem to be responsible for a broad range of different memory
effects. Each effect has it's own envelope of temporal characteristics, that include onset time, rise-time, and decay time. Many of these mechanisms have been discovered relatively recently, but general knowledge of them, along with an understanding of their diversity, has been forming for over three decades.
. . . . . . .
In Artificial Neural Networks
Learning, in artificial neural networks
(ANNs), is (traditionally) the process of adjusting a set of weight values
to bring a neuron
's output response closer to a desired response for a given set of inputs. The weight values
represent connection strengths between neurons
. . . . . . .
Learning Is Ubiquitous
Learning seems to be always occurring. That is, it is a phenomenon which can be counted on to occur, like chemical interactions, or gravity. While it is just speculation, it seems to have a functional and structural characteristic of many small, relatively fast events contributing slowly to learning (e.g., an explanatory metaphor
). In this regard, learning also seems to be a constituent (trait?/characteristic?/attribute?) of consciousness
. . . . . . .
Learning's Relationship to Consciousness?
Learning, —that is, not merely reacting
to stimuli, but adapting
future responses to similar stimuli— seems to be intertwined with consciousness
Many philosophers of mind, including the one I admire most, David Chalmers
, eschew the use of reductionism to better understand consciousness. They claim that reductionism is not useful in this regard because consciousness is not reducible into constituent parts, or because it IS the agent that understands, and so can not be reduced to some other things capable of promoting understanding.
I can appreciate this argument, and can't really argue against it in any authoritative way. My armchair arguments, however, go something like this: Isn't insisting that reductionism not be used for analysis of a marginally understood entity, ITSELF a form of reductionism? — and also this: Nobody has ever seen the high-energy particles that fly through cloud chambers. In fact, we have not even seen their direct effects on other things. The observed effects are at least twice-removed from the unseeable particles and velocities that cause them.
Still, we are able to study those constituent causes and learn a great deal about them from observing the effects that ionized gas has on water vapor. Also, the eyes of the inventor/innovator in all of us will light-up any time people limit their own inquiries based on authority-driven rules-of-thumb, or "conventional wisdom." When things "like" this ("of this nature?") are evident, it is human-nature to want to explore the underlying assumptions for ourselves. Maybe it's not normal
human-nature, but whatever the DSM might say it is, I'm certainly guilty of it.
. . . . . . .
- Determining Similarity — Associating
The process of Grouping similar (like) things together may seem fairly simple at a glance,
but it is actually quite complex and nuanced. Much like consciousness itself,
one of the reasons similarity is hard to understand is because the concepts
and mechanisms at the heart of similarity are elusive. That is, they often can't be rigorously expressed.
Determining what constitutes
"similarity" in an implementable way, is where this hard-problem is exposed. (e.g., rhyming words). This is
also embodied in "discrimination" and "discernment." While it can be relatively simple (grouping by visual
morphology such as shape and color), it can also quickly go from simple to hard at more abstract levels. Consider
things such as rhythm/duration-patterns which are often easily discernable algorithmically. Now consider things like
rhyming, similar movements and behaviors, historic comparisons which are often
said to rhyme rather than repeat, etc., That said, explaining things like rhyming and things like the
sentiment in the phrase "something in the
way she moves" in an implementable way, it turns out, may not be as simple as it sounds.
To recap: The determination of what constitutes similarity is where the rubber meets the road. That is, it is the
"hard problem" part of grouping similar things. It can be extremely complex and nuanced. It is an elusive
problem, in that sometimes we recognize similarities without even being able to consciously perceive or
explain how the things are similar.
Consider how we are able to construct and understand metaphors, as one (relatively easy to grasp)
example. The similarity between, say, "sharpness" and a particular cheese-taste experience is a similarity
that might be completely inexpressible in words, but for the metaphor of using the word "sharp" to describe
a flavor and mouth feel. Rhyming, also, is clearly something we experience as a similarity, as are
genres of music and other art forms. We sometimes relate things that can't be concisely described by
using metaphors or analogies (e.g., coffee smells a little like chocolate tastes). The point
here is that the key to understanding this, and to how
we group alike things (experiences? sensations? concepts? relationships? interactions?) together is
in understanding the complexity of determining what constitutes likeness or similarity.
There is even a meta-level to this. Strategies for determining how different ways of grouping like
things together can, themselves, be grouped based on their (often hard to fully grasp) similarities
and differences. In this first-order grouping, the concept of "similar" itself, can be grouped with
other, "similar" concepts, such as: tangential, related, class, phylum, division, etc.
Less obvious might be contextual opposite, in which pairs such as: prince-pauper, prince-tyrant,
prince-princess, and, to make the point, cynicism-optimism are all similar relationships in that they
are logical opposites. Interestingly, cynicism-optimism can also be grouped with similar things where
both assume a given outcome before having evidence to support it — in this case, one assumes a positive
outcome, and one assumes a negative outcome but both assume an outcome. That can also be grouped into a set of
"things that are opposites
that are also the same." Not sure if this meta-layering ever reaches a terminal (top-most order) state, but
it seems like the brain's incessant and unyielding drive to find each next-level state may have something to
do with consciousness.
- Determining Difference — The ability to perceive differences between things and situations.
A ping-pong ball and the moon are both spheres. That's one way in which they are similar. The differences
between them include size, location, and even purpose, as well as many subtle differences.
That we are able to
discern, say, the difference between a sardonic smile and a sarcastic smile is a testament to the power
of a system that is able to adapt it's future responses based on current experiences. Like the ability
to perceive sameness, the ability to perceive difference includes some very subtle and nuanced underlying
connotations. The hard problem here, is in determining just what it is that constitutes a "difference." What is
it about the sardonic smile that makes it different from the sarcastic smile. What makes an open, unassuming
smile different from these two types of smiles. In this context, what is meant by type? It is, perhaps, a
dichotomy that discernment is an underlying mechanism in both determining similarity, and determining difference.
- Adapting — The process of Learning
Inter-adaptation —Not simply responding in a preprogrammed way to a given stimulation
(interacting), but altering future responses to the same (or similar) stimulation— is enabled by
processes within neurons that are able to form and decay without needing to sacrifice the entire cell.
The changes in these supporting processes to
future responses can be triggered by dissonance detected in current
or by similarities and difference determined/detected, in part, by the above discussed functionality.
Future, in this case, may be measured in fractions of a second, to years, or even centuries. The dissonance
itself may be caused by differences between current stimulus and previously impressed (learned)
Inter-adaptation of this nature occurs, and is observable —obviously— between humans and other
animals, but there is also very strong evidence that this inter-adaptation occurs, even at a sub-atomic
level. The double-slit experiment seems to demonstrate inter-adaptation, even between
high-level animals such as human observers, and sub-atomic particles. Though in the world of
thought-experiments, perhaps not cats.
Dissonance — randomness, confusion, chaos, fast oscillation(instability?), lack of balance,
inconsistency (can you
see the similarity?) seem to be triggering factors in adaptation as
well as in determining similarities and differences (i.e., determining ways in which things can be
considered similar or different to each other and to things in our past experiences). When there is dissonance, there
seems (at all levels) to be a need to make things more balanced, to make the sensory information being experienced
more consistent, less confusing, more explainable in light of previous experiences. Offhandedly, there is a desire
to make the experienced sensory input make sense.