Post by jmfbahcivPost by TomassoMuch of that stuff in your list is application of physics, not application of
representation. Some of it is about compression. Some is about
discretisation.
If a representational framework is a reasonable for holding models of
parts of reality (whether RDF or anything else) then information can
stop being measured in "bits" and start being measured as "answers to
questions worth answering".
Noto Bene, the above sentence is not trivial. Furthermore, it is not well
understood by many people. It's starting to find a home in fields like
"sense making".
If that sentence is not trivial, would you please edit it so that
it's readable? Then I'll be able to analyze what your meaning.
(1) Information is a measure of the change in a probability distribution
of a receiver as a results of a sender sending a signal to the receiver.
This is measured in bits, as developed by Shannon and Weaver, and
others.
(2) The probability distribution is specific to a particular belief, and
the probability of it being true (in the sense of agreeing with observation
and being able to make predictions about reality, or in the sense that
the sender intends the receiver to modify a belief).
(2a) In carrying out experiments or observational studies, the concept
of "sender" is extended to carrying out tests or measurements.
(3) The specificity of the probability distribution is about the statuses
for which the receiver held, or holds probabilities, as well as the
probabilities themseves. We can call these statuses a model space,
and the probabilities give likelihoods of particular models being
believed, or true, or useable for prediction, etc.
(4) The notion of framework is the structure of the model space.
This is composed of terms, constraints, and context or semantic
associations with other frameworks. RDF is a part of XML for
defining these (RDF = resource definition framework), and is a
way to denote a model space. It is not the only way, but is one
way.
(5) The notion of importance, or 'questions worth answering' can
be part of a framework, but not in the probabilities in the probability
distribution, but in the surrounding context. This includes the
stances, intentions and motivations of the receiver.
(6) Within this kind of framework, "bits" are about the influence
of the particular incoming signals. This is a lot different operationally
from "bits" when used to characterise parts of bytes and Megabytes,
or in lossy compression, or self-assembling robots, etc.
(7) Within this kind of framework, an incoming signal may result
in not merely changes to probabilities, but a revision of the framework
and the contexts in which it is embedded. Ie, structural changes as
well as simple tweaks of beliefs.
(8) For example, a knowledge representation framework may get
changes, or something which was held to be important may become
unimportant, or uncertainty may increase after a signal is received.
In the latter case, a seismic upheaval of the framework is anticipated,
but may take some time to eventuate.
Feynman famously asked "how do you know that"? His question
required a detailed answer. A second question of this kind is
"what it it that you know"? This is a question about model spaces
and frameworks, and also requires a detailed answer. It is a core
part of AI. Deep down, it is also part of physics.
Tomasso.