Ricardo Ribeiro Gudwin wrote:
> Don Mikulecky wrote:
>
> > Don mikulecky replies:
> >
> > Ricardo Ribeiro Gudwin wrote:
> > > > please define "levels of resolution"
> > >
> > > I don't have exactly a formal definition (maybe I should have one), but
> > levelsof
> > > resolution give respect to levels of functional details when describing a
> > model.
> >
> > now you asked me some specific questions about Rosen's work and these terms.
> > how am
> > I supposed to answer?
>
> :-) Ok, Don, I see your point ! I'll try to do it better later on this
message.
>
> > > The best example is the following ... at a higher level of resolution, we
> have
> > a
> > > globe, with all the continents. In a lower resolution, we have the USA
map.
> >
> > You said:I partially agree with you here, Don ! We may have "distinct ways
of
> > interactingwith
> > a system" that are not "different levels of resolution". For example,
consider
> > that complex system we use to call a VCR (the real one). We may see it as an
> > artifact to convert a movie from its magnetic form to the television, OR as
a
> > good niche for insects, as Alexei already pointed out earlier. These are
> > distinct
> > ways of interaction that do not pressupose levels of resolution.
> >
> > I find this VERY confusing. i'm talking about ways of interacting and you
> bring
> > up
> > this idea which you can't really define.....how am I to relate them?
>
> OK, this is time for your revenge. :-) Let's try to put it on again in other
> way.I
> have a particular example of a complex system. Just to put it on, this system
> is the physical artifact: VCR. I have been given this artifact for the first
> time,
> and I don't know how does it work or what can I do with it. I start analyzing
> this complex system, using the reductionistic methodology. I start interacting
> with it, putting a cassette inside it and pushing some buttons, I got an image
> in my TV. This is one way of interaction. Suppose now that I am an ant and
> I see the same complex system. I try to interact with it, and see that if I
> put my eggs inside it, they become warm and grow well. Suppose now
> that I am a salesman and see this complex system and start interacting
> with it. I pay very cheap for it and can win a good money with it, because
> it sells very well. Suppose now I am a technic in electronics and I like old
> VCR's. I start interacting with this complex systems, and I discover that
> it is full of old chips that are not factored anymore. So I discover a good
> source of old chips. And I could invent other examples ...
> Notice that all those interactions gives a different description of our
> complex system, based on the type of interaction was performed.
> Despite being different types of interaction, they do not correspond
> to different levels of resolution. They are somewhat parallel's points of
> view describing partially our complex system by different angles. Now
> I will start an example were different levels of resolution take on.
> Let's take the same VCR. I start interacting with it and can decompose
> it into mechanical parts and electronic parts. Then I start interacting with
> the electronic parts (the boards), and try to decompose them into
> components and wires. I look at the wires and I can decompose them
> into plastic and copper. Notice that this time I interacted with the complex
> system, but within different levels of resolution. Maybe the trick here is
> that I am analyzing the same functionality (the ability to exibit images on
> the TV). What I was trying to say is that I can perform different interactions
> WITHOUT going through different levels of resolution. Despite I can perform
> different types of interaction WITHIN different level of resolution.
> Does it became more clear now ?
>
yes very much
> > > In a lower resolution, we have the map of Maryland. In a lower resolution
we
> > > have the map of Baltimore. In a lower resolution we have the map of
> > > dowtown Baltimore, and so on. Notice that details in each level of
> resolution
> > > are missing on the higher resolution model.
> > > If you want to get a more of this, I recomend:
> > > Meystel, A.M. - Intelligent Systems : A Semiotic Perspective,
International
> > > Journal of Intelligent Control and Systems, vol. 1, n. 1, pp. 31-57, 1996
> > > Maybe you may want to see also:
> > > http://www.isd.cme.nist.gov/projects/semiotics97/whatis.html
> > > and
> > > http://www.dca.fee.unicamp.br/~gudwin/semiotics/semiotics.html
> > > that is a copy of a page that disapeared from its original site (also from
> > > Meystel),
> > >
> > > analyzing the intereaction between semiotics and intelligent systems.
> >
> > Oh, but you are the expert on these things! Why are you asking me to go
> > elsewhere?
>
> Wait a minute ! I am not saying to you ... go and read 4 or 5 books anda
hundred
> of
> articles in magazines. I am giving you 1 article and 2 web pages !
> But seriously, ... , :-) , ... , (I can't stop laughing), ... , I got your
> point. I
> am
> getting the most important definitions and concepts and building a page with
> a summary on Computational Semiotics in my Web site. You are right.
> It is not in 5-10 lines that you will explain complexity.
> ok...we are getting somewhere......I look forward to the pages....
> > To quote you again:
> >
> > "what I want now is to understand EXACTLY how this formalization looks like,
> > and differentiate it from the formalization of a mechanism in my mind. "
can
> I
> > ask
> > the same?
> >
> > further quote:
> > "This is not a matter of lazyness to study
> > or whatever, but that a good answer to our questions from you can save hours
> of
> > reading. I confess that I don't have enough time by now to dedicate to this
> > study, but I also don't want to miss the point into the discussions. This is
> why
> > I keep asking."
> >
> > If you aks this of me...why do you send me elsewhere for my answers to your
> > questions.....it seems a bit asymmetrical?
>
> Well, this is not exactly the same thing. To go and study Rosen I first have
> tofind
> its publications (some are easy to find, others not), and I just got to
> myself that I will have to do that sooner or later. But I have to reserve a
long
> time for doing so. What I am asking you to do is to click the mouse onto the
> references I gave you and read it ! THIS is asymmetrical !
> But, you are right. I will try to put it out on my web site.
>
> > > > > Actually, to be more precise, in Computational Semiotics we would
refer
> > > > > toRHEMATIC
> > > > > KNOWLEDGE, which includes senses, objects and relations. Then,
> > > > > a RHEMATIC KNOWLEDGE may be a SENSORIAL KNOWLEDGE, a
> > > > > KNOWLEDGE OF OBJECTS and a RELATIONAL KNOWLEDGE (also
> > > > > called KNOWLEDGE OF OCCURRENCES). So, we have three types of
> > > > > rhematic knowledge.
> > > >
> > > > to help me unerstand...lets use the number system as an example...how
does
> > it
> > > > fit
> > > > into the above?
> > >
> > > I believe that I have an example that may allow you to better understand.
> >
> > The case that allows direct comparison of the two approaches is the number
> > system.
> > Could you do it for that example please?
>
> I'll try ... keep reading ...
>
> > > Roughly
> > > speaking, you may associate a rhematic knowledge as the semantic
> > > that can be assigned to single (isolated words) in a natural language.
> >
> > Like the symbols for numbers?
>
> Not only symbols. I may use symbols, indexes and icons.A symbol would be like
> "2",
> using the indo-arabic graphology.
> An icon would be more like the romans "I, II, III" - notice how they
> mimic the number they represent (at least from 1 to 3, the 4 (IV) is not
> an icon anymore, but a combination of icon and symbol)
>
> > > This is
> > > not a definition, ... , just a way of understanding what it is. The full
> > > definition is far more involved.
> >
> > Oh, but you can clearly summarize it for me in a few words and save me
> > trime...as you seem to expect of others?
>
> You really don't loose a chance to pull the carpet on me. :-)(there is such
> expression in english - "pull the carpet" ?)
>
yes...I am nasty...sorry!
> > > Again, we may make the following associations:
> > >
> > > SENSORIAL KNOWLEDGE - the semantic of adjectives in a natural
> > > language
> > > KNOWLEDGE OF OBJECTS - the semantic of substantives in a
> > > natural language
> > > RELATIONAL KNOWLEDGE - the semantic of verbs in a natural
> > > language.
> >
> > I don't see that these definitions apply OUTSIDE of language. what about
the
> > number system?
>
> OK, you win ... ! I was trying to avoid the mathematics, but let's go. I will
> tryto
> summarize, but the complete developments are in my articles (in my
> Publications page, if you want to read them). Let's first understand the
> connection to language. To each term in language, we (humans) are able
> to make a "ground". For example, for the word "red", I can relate to
> a perceptual experience (the experience of red), that sustain that concept.
> Suppose now that I want a robot to understand what is "red". I will have
> to get a color camera and point to something red and say "red" to the
> robot. Then it would associate the sound of the word "red", to the color
> red. Suppose now that in my camera, the pixels are described in terms of
> its (R,G,B) coordinates. So, a whole image will be an array of triplets
> RGB. You may see this as a policomplex number (or structured number).
> Now, each image is one structured number (I will write s-number in order
> to avoid too much typing). Each image like that is a "rhematic iconic
> sensorial specific knowledge unit". I can show different images with
> the color red, and then I will get a set of s-numbers. This would be a
> "rhematic iconic sensorial generic knowledge unit". Then, the robot
> start walking along my red object, collecting s-numbers. Then, it
> starts to conjecture why every time it points the camera to one place,
> it gets this same s-number (or correlated s-numbers, with a great
> emphasis on their R componenet). It then generates a hypothesis that
> this is due to SOMETHING that is at a fixed coordinate, that has the
> property of being red. It then discovered an OBJECT in place (x,y).
> This object, is described in terms of a set of atributes, e.g., color,
> coordinates x and y, shape, etc. To allow the description an object,
> I use a "mathematical object" (I can't put the mathematical definition
> of a mathematical object here, ... , it took 4 or 5 pages full of equations
> in my PhD thesis - see my references). But, just to summarize, a
> mathematical object is a special kind of set, built with s-numbers and
> some tricks using functions and relations. So, it is still a mathematical
> entity. Objects are then a creation of our mind, that comes in two
> flavours: "rhematic object specific knowledge unit" and "rhematic object
> generic knowledge unit". The specific object is a particular instance of
> a class of objects that we assume to exist in the world. The generic object
> is a set of correlated objects. I use a mathematical object to describe both
> of them (again, with some tricks). With sensorial KU and object KU, I
> would be able to describe a static world. But, very soon our robot
> realized that objects are not static, and that their atributes changes with
> time. This change is modeled through the "rhematic knowledge of occurrences",
> or rhematic relational knowledge units, that again came in two flavours,
> specific and generic. For example, to understand the semantic of the verb
> "to warm". It pressuposes an object, with an atribute "temperature" that
> was first "low" and later "high". I can model this type of information through
> the use of another mathematical entity, the meta-object (again, 5 or 6 pages
> full of equations to define them). These meta-objects are also special
> kinds of sets. Now, how do I translate from language to mathematics ?
> I just got rhematic sensorial KU's (adjectives), and associate to s-numbers,
> rhematic object KU's (substantives) and associate to mathematical objects,
> and rhematic relational KU's (verbs) and associate to meta-objects, in the
> sense that s-numbers are models of adjectives, m-objects are models to
> substantives and meta-objects are models to verbs. But this is not all,
> ... , I still have dicent knowledge, that is, the knowledge about the truth
> in the real world, and argumentative knowledge, that is the knowledge
> of transforming knowledge. Dicent knowledge is made with the combination
> of rhematic pieces of knowledge. Argumentative knowledge came usually
> in three flavours, "deduction", "induction" and "abduction". These are
> terminologies for special kinds of functions manipulating knowledge in
> a mathematical form (given by s-numbers, m-objects and meta-objects,
> when working with rhematic knowledge). Using these functions, we
> can make manipulations with the knowledge the robot first acquired,
> generating its future behavior. This explanation is really more confortable
> with some images ... see my web page to better get these concepts in.
> Did you have stomach for reading until here ? Was it understandable ?
yes and clearly we are converging..this is exactly the kind of thing Bill Dress
seems to be trying to do.
> Please feedback me. I am always trying to put these concepts in a
> more digestible format. They are not easy to get (I believe the same
> happens with complexity).
> ok/...I'm printing this off and taking it on vacation with me...tell you more
the
> week after next.
> > > Notice that these are only rough ideas in order to allow us to understand.
> > > Sensorial knowledge is more than simply the semantic of adjectives.
> >
> > In chapters 2-5 of "Anticipatory systems" Rosen does a thourough job of
> dealing
> > with
> > these issues using the modeling relation. I hope to match these...can you
> help?
>
> I believe we have one copy of this book in our library. I will try to get
itnext
> week and take a look.
>
> > > They
> > > include all perceptual information that comes directly from sensors, or
> > > are generalizations of such information.
> >
> > yes...sounds very much like our percepts
> >
> > > It may also include emotions and
> > > actions. Knowledge of objects is the knowledge of the singularities we
> > > may extract from the world - the models we may get from the world.
> >
> > ok...consistent with the anticipatory systems concept
> >
> > > Everything that we could assign a name Relational knowledge gives
> > > respect to the possible relations that objects may have each other, in
> > > terms of correlations or either causal relations. If you want more
> > > detail, please see my publications on the theme, specially the two
> > > reports at the end of the WWW page:
> > > http://www.dca.fee.unicamp.br/~gudwin/Publications
> > >
> >
> > ah...but can't you give me the information more clearly here?
>
> Just for one reason ... I will need graphics, and unfortunatelly itis not
> everybody
> that can read messages with graphics. So, we
> are not expected to put messages with graphics to a list. But
> I am improving my web site. Take a look and feedback me.
> (I got you this time, ahn.. , didn't I ? :-) )
will do....touche
>
>
> > > > I suspect this differentiation may have hidden baggage in it....one big
> > > problem
> > > > Rosen encounters is that the old categories (even in semiotics I
suspect)
> > were
> > > > made
> > > > to conform to the Newtonian paradigm...this may be one case.......
> > > >
> > > > why is it important?..because relational things as in relational models
> are
> > as
> > > > much
> > > > objects as a tree is...yet they get expressed as mappings in category
> > > > theory..,.the
> > > > beauty of category theory is that it allows us to treat mappings as
> objects
> > as
> > > > well
> > > > as the way of relating sets
> > >
> > > This is why I am trying to understand better Rosen's ideas. I would like
> > > toevaluate
> > > this too.
> > >
> > > > > believe that the word "percept" will
> > > > > better suit what I call SENSORIAL KNOWLEDGE, but if Roses use it in
> > > > > the broad sense of RHEMATIC KNOWLEDGE, then it is OK !\
> > > >
> > > > please define rhematic knowledge..my sources on semiotics don't
> > > >
> > >
> > > Already done in some paragraphs above.
> >
> > ???????????? am I missing something?
>
> Rhematic knowledge comprises the pieces of knowledge that can be assigned
> toisolated
> words in a natural languages. So, they are symbols pointing to icons that
> represent isolated components of a world.
>
> > > I differentiate rhematic knowledge,i.e.,
> > > the
> > > semantic of single terms from dicent knowledge (the semantic of
> > > phrases, or composition of terms) from argumentative knowledge, i.e. the
> > > knowledge of reasoning procedures, that will include deduction, induction
> > > and abduction.
> >
> > Lots more terms which I have not encountered in scientific discourse
> > before...please forgive my ignorance. Can you say it in "common sense"
> > language which you know so well?
>
> Dicent knowledge - the knowledge of truth in environment. They are made
> ofrhematic
> knowledge structured coupled, to which is added something called
> a "truth value". It can be represented again by a mathematical entity.
>
> Argumentative knowledge is the knowledge of manipulating knowledge.
> It comes in basic three families: deduction, induction and abduction.
> Deduction (this one I believe you have already heard before) is associated
> to knowledge extraction. For example, you already know f(x) and then for
> x = 5 you extract f(5).
> Induction is associated to knowledge generation. For example, you have
> a set of points (x,y), and based on them you generate a spline function
> y = f(x).
> Abduction is associated to knowledge selection. For example, you have
> the same set of points (x,y), and then you try to fit different types of
> curves, e.g., a parabol, an ellypsis, an hiperbolic function, a sin, etc.
> After testing them, you decide, e.g., that the courve is a polynomy.
> You may aply these methods to the mathematical entities describing
> the other types of knowledge, generating new pieces of knowledge,
> successivelly, leading to a sequence of interpretations that can be
> associated to the stream of thought.
>
> > > > > If it is not all there, then what exactly is missing ? What would be a
> > > > > FORMAL DESCRIPTION for a complex system ?
> > > >
> > > > herre's the rub!......a complex system has more than one formal
> > > description...in
> > > > fact an infinite number of mechanistic descriptions can not catch
it...due
> > to
> > > > the
> > > > semantics it contains....non-computable elements
> > >
> > > Here again this problem of infinite. We humans are able to understand
> > infiniteby
> > > performing induction. We just compress information and then we have
> > > a formal definition for infinite. An interval is a representation for an
> > > infinite
> > > set
> > > of points. What I am trying to investigate here is if will be something
like
> > > that
> > > for complexity theory. You say, ... , there is an infinite number of
formal
> > > descriptions,
> > > but isn't there a way of compressing this infinite number into something
> > finite
> > > ?
> >
> > yes...we did that hundreds of years ago..it's called the Newtonian
> > paradigm...sometimes refered to mechanistic science or
> > reductionism.....unfortunately...it misses all the rest ....as will any
> partial,
> > finite set...however..for many purposes, people find this adequate
>
> One thing doesn't fit in my mind ... how do WE HUMANS are able toencode such
> knowledge ? If we are able to encode it in our brain, then
> THERE SHOULD BE a mathematical representation for it.
>
> > > Like infinite numbers in an interval, ... , like infinite sums to
calculate
> a
> > > sin(x)
> > > ?
> > > This compacted, compressed piece of information is what I am trying to
call
> > > here a FORMAL DESCRIPTION for a complex system.
> >
> > yes...that ois what I understood.....and you may get away with a single or
few
> > such
> > descriptions if your needs are sufficiently circumscribed...however you'll
> never
> > see
> > the entire system that way
>
> Again, ... how do we encode it in our brain ?
>
> > > > there is more common sense to this than you acknowledge...why should all
> > > aspects
> > > > of
> > > > the world be formalizable...why do you ignore Goedel?
> > >
> > > The question is ... if we (humans) are able to understand something, there
> > > shouldbe
> > > a mental model for such something.
> >
> > "understand" here is poorly defined....however what you say is
> true.....however
> > as
> > we manufacture models(causal event--->natural system<--encoding and
> > decoding------> formal system<--implication) [should be two arrows one for
> > encoding one for
> > decoding]
> > we find some are not really models...they don't commute...only when the
> diagram
> > commutes does it make a model.
>
> Can you explain a little bit more on this ... it is still a little bit confuse
> to
> me.
yes...the diagram (graphics) is not possible...tried to do it but so
clumsy....I'll
try again when I get back.....if you want try this:
http://views.vcu.edu/~mikuleck/chaos2.htm
go to the section on the "Modeling Relation"
>
>
> > > If there is such mental model, then there
> > > should be a corresponding formal model also.
> >
> > as you can see we are in a semantic bind.....in the modeling relation
> > above.....which when it commutes is a model...the formal system is clearly
at
> a
> > specific place. This ensures that we realize that the encoding and decoding
> DO
> > NOT
> > COME FROM THE FORMAL SYSTEM. Thus a maodel may become a natural system to
be
> > placed
> > on the left side of the diagram...THEN a formal system may be found to
> > correspond to
> > it (new encoding and decoding) and if this new, higher level diagram
commutes
> we
> > again have a model.
>
> I see this differently ...natural system -> enconding -> formal system (our
> brain)
> -> decoding -> natural system
> I still don't understand what you said.
>
> > > The only thing we cannot
> > > formalize is what we cannot understand !
> >
> > then you deny Goedel's proof and the experience with number theory?...sorry
> then
> > you
> > are wrong, I'm afraid.
>
> I believe that we are talking about different things.
>
> > > I believe that one problem is our understanding for the word "formal".
> > > You seem to use it as a synonim of "syntactic". I use it as both
"syntactic
> > and
> > > semantics".
> >
> > usually the semantics is in the encoding and decoding...especially in
decoding
> > where
> > meaning is assigned..
>
> I understand the words encoding and decoding in the following way ...We have
an
> external space of representation (the environment, the
> real world), where some signs take place. These signs are interpreted
> through our sensors, creating other signs (their interpretants), in
> an interpreter's internal representation space. This is encoding.
no here is the problem...we call this the formation of a percept......then this
is
matched with a formal explanation...that's the encoding
> These signs
> taking place in the internal representation space will trigger a
> massively parallel chain of interpretations within the internal
> representation space, until there is a prescriptive sign that
> leads (through our actuators) to a new sign on the external
> representation space. This is decoding.
we call decoding the attempt to match the result of manipulating the formal
system
with the observed causal event in the natural system.....the idea of
entailment...causality is central here.....we are describing mental processes
meant
to decipher the external world
> Once a new sign is
> created on external world, it is able to be interpreted to
> another interpreter. The signs holding within the internal
> representation space can only be interpreted by the same
> interpreter. This is a summary description of a system
> performing semiosis.
> I am not sure you assign the same meaning to the words
> encoding and decoding.
> yes..ours are different...I'll rovide aq translation eventually...(more
#$@^^%$
> work!)
> > > The great difference from what we are doing in Computational
> > > Semiotics and what was made by people from Artificial Intelligence is that
> > > AI used only syntactic models, and we are trying to use models that
consider
> > > both syntactics and semantics. Our symbols are not empty like those in AI.
> > > They are full of meaning. And we don't use only symbols, we use also icons
> > > and indexes.
> >
> > you can't succeed by trying to turn semantics into syntax
>
> I am not trying to do this ! This is what I am trying to show you !I am fully
> considering semantics, because I ground it into the sensorial
> information I get from environment.
good I'll get it yet....I need to study this now
>
>
> > > > and I would challange you to tell me (common sense question) how
something
> > > > computational can be intelligent??????
> > >
> > > The first idea when we used this name "Computational Semiotics" was
> becausewe
> > > were
> > > trying to emulate the process of Semiosis within a computer. Maybe
> > > we may have to reformulate this name after the (possible) impact of
> complexity
> > > into CS. The bigger claim here is that if something performs semiosis,
then
> it
> > > is
> > > intelligent. There are different levels of intelligence, each level
related
> to
> > > the
> > > number of different types of knowledge it is able to process. Simpler
types,
> > > lower intelligence ... greater number of types, higher intelligence. We
may
> > > either say that at a certain level, mechanistic systems may achieve a
> certain
> > > level of intelligence. Actually this is what I am currently trying to
> > consider.
> >
> > ok...thanks for the clarification
> >
> > I detect places where Bill dress' paper could be a huge help in bring these
> > things
> > to a sharper focus. Thanks for your help.
> > Don
>
> I will write to him asking for a copy. I really want to read it.I hope I could
> make
> myself a little bit clear this time. If not, just say me and I
> will try again (if you think it is worth the value).
> Hope the wind doesn't give you too much damage.
> Best regards,
> Ricardo
>
> --
> //\\\
> (o o)
> +-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-oOO--(_)--OOo-=-=-+
> \ Prof. Ricardo Ribeiro Gudwin /
> / Intelligent Systems Development Group \
> \ DCA - FEEC - UNICAMP | INTERNET /
> / Caixa Postal 6101 | gudwin@dca.fee.unicamp.br \
> \ 13081-970 Campinas, SP | gudwin@fee.unicamp.br /
> / BRAZIL | gudwin@correionet.com.br \
> +-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-+
> \ URL: http://www.dca.fee.unicamp.br/~gudwin/ /
> / Telephones: +55 (19) 788-3819 DCA/Unicamp (University) \
> \ +55 (19) 254-0184 Residencia (Home) /
> / FAX: +55 (19) 289-1395 \
> +-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-+
thanks!
we are getting somewhere!
Don