Hi Ray,

Let me top-post a hot take. As such, it may be off-target, miss the point,
and be randomly disconnected. etc. (and I sometimes manage to word things
in an offensive manner, if so, apologies in advance; I mean to be direct).

Goguen, 'Realization is Universal', seems to be from 1972; I did not find a
PDF, but I did get https://core.ac.uk/download/pdf/82200864.pdf Nerode -
Universal Realization (1979) and This:
https://www.cambridge.org/core/journals/mathematical-structures-in-computer-science/article/categorical-manifesto/89D0DC6DDF8A8176522AEF1450AD5E54
Goguen - A categorical Manifesto (1991)

So ... these are old, dated, maybe hopelessly dated. Not in a bad way, but
in a good way, actually. Category theory was considered to be obscure
through the 1980's; category theory applied to comp sci was just arcane
black art. That has changed in the last 3 decades. -- We've got more than a
few programming languages -- ML, CaML, Haskell, F# that are explicitly
categorical, and they're not even obscure, people actually write commercial
applications in them... so the world has changed.  There are at least a
couple of books that teach how to program in ML/CaML by teaching category
theory (I don't have titles/authors; spotted them in the university
library)  I'm quite certain I saw a development of "realization is
universal" in chapter 2 or 3 or 4 in one of these books.

There's even this:  http://www.tac.mta.ca/tac/reprints/articles/22/tr22.pdf
(1998) and this: http://www.cs.man.ac.uk/~david/categories/book/book.pdf
(no year given) and I'm fairly certain  that they recap what Goguen and
Nerode were trying to say years earlier. And they probably say it more
clearly. In the sciences, it pays to read the most modern stuff available.

------
Moving on to "Plans and the Structure of Behavior", by Miller, Galanter,
Pribram (1960) wow .. that's ... really old. It predates the first AI
summer, when "planning languages" (for motion and navigation and action)
and "expert systems" and 100 other flowers blossomed. Vast quantities of
stuff was discovered in the subsequent decades, uniting computation, logic,
language, syntax, constraints, categories and types, and forming a bedrock
of modern comp-sci.

So sure .. at a certain abstract level, motion and behavior are structured,
and can be realized via state machines or monoids or more complex
algorithms .. and this is definitely fun to read about and understand ... I
mean, I did, and I liked it ... but ...

.. but how useful is it? For example: a famous paper from microsoft shows
that SQL and noSQL databases are category-theoretic duals to one-another.
Knowing this sharply clarifies one's thinking, and also lets you sleep easy
at night, as you no longer have "fear of missing out" (i.e. of using the
wrong technology)

It's also folk-knowledge that functions returning error codes are
category-theoretic duals to try/catch exception mechanisms. Helps with
sleep at night, and also helps design high-performance code.

Garbage-collection is category-theory dual to reference counting. (Java vs.
C++ smart pointers, in real life.)

Monodial actions are category-theoretic adjoints to finite state machines
... (I think this close to what "realization is universal" is about!? Not
sure.)

Types, categories and languages are locked in with one another (this is one
deep fundamental reason why I keep nattering on about "link grammar")

I do NOT know of any dualities involving motion-planning; surely some of
these must be known; I'd love to actually hear about them.  Searching for
them in the 1960's and 1970's is the wrong place to look, though. Look into
the last 10-20 years, instead.

The exciting place to hunt for dualities is in neural-nets vs. sheaves,
which is where I'm looking.

So I dunno. Do plow into the two PDF's I gave above... they give a
foundational cornerstone. Then, whether there are category-theoretic duels
involving motion planning... let me know if you find any.

-- Linas

p.s. totally off-topic, but I recently found this to be kick-butt:  "Forced
moves or good tricks in design space?  Landmarks in the  evolution of
neural mechanisms for action selection", Tony J. Prescott    (2007)
https://www.academia.edu/30717257/Forced_Moves_or_Good_Tricks_in_Design_Space_Landmarks_in_the_Evolution_of_Neural_Mechanisms_for_Action_Selection

On Thu, Feb 27, 2020 at 5:37 PM Raymond T. Melton <[email protected]>
wrote:

> Linas Vepstas:
> Belated Thank You for the reply.
> By "structure of behavior", I have in mind an application of the type of
> results found in a paper by Goguen, 'Realization is Universal', wherein he
> shows that behavior of machines is adjoint to minimal realization.  While
> reading "Plans and the Structure of Behavior", by Miller and others, their
> description of Plans versus Images seems to describe a duality.  And
> adjoints are sometimes as close as you can get to duals.  My question then
> is really, Has anyone here chased this idea down, tried to apply it to
> motion planning, and shown that it must be a dead end?
> If not, I'm going to run with it.
>
> Regards, Ray.
> P.S. Oops, sorry for top posting.
>
>
> On Thu, Feb 13, 2020 at 12:00 PM Linas Vepstas <[email protected]>
> wrote:
>
>> Hi Ray,
>>
>> On Thu, Feb 13, 2020 at 11:46 AM ray melton <[email protected]> wrote:
>>
>>> Is anyone else here working on planning, and the structure of behavior?
>>> As in the planning of motion, say? Generating such plans?
>>>
>>
>> Not at this time, not that I know of. What do you want to do? What's
>> wrong with stock, off-the-shelf motion planners for quadricopters or
>> rolling bots or whatever?
>>
>> I am interested in in attaching natural language to motion, i.e. to issue
>> commands "go here, go there" or to provide directions: "go to the first
>> light and then take a left" or to answer questions about navigation:
>> "what's the best way to get from here to there?" -- well, I guess that
>> sounds silly, because "siri already does it" so I'm more about "how do I
>> talk about space and motion using natural language".  ... well, I already
>> have some fairly fixed ideas on how to do this, just no forum in which to
>> explore them.
>>
>> If by "structure of behavior" .. hmm, yes, there are also some very
>> interesting things about motion of football players (or motion of military
>> troops, or protesters, or Mexican border-crossers, if we want to get
>> Orwellian) and how those motions are strategized and planned.  Worked on
>> that once, but there's a huge gap between what one can imagine, and
>> existing software.
>>
>> --linas
>> --
>> cassette tapes - analog TV - film cameras - you
>>
>> --
>> You received this message because you are subscribed to the Google Groups
>> "opencog" group.
>> To unsubscribe from this group and stop receiving emails from it, send an
>> email to [email protected].
>> To view this discussion on the web visit
>> https://groups.google.com/d/msgid/opencog/CAHrUA34A-%2BPVELM4DUT9xFQFK4UJCPq0yLmUjZKP_HdKDVAj%2Bg%40mail.gmail.com
>> <https://groups.google.com/d/msgid/opencog/CAHrUA34A-%2BPVELM4DUT9xFQFK4UJCPq0yLmUjZKP_HdKDVAj%2Bg%40mail.gmail.com?utm_medium=email&utm_source=footer>
>> .
>>
> --
> You received this message because you are subscribed to the Google Groups
> "opencog" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to [email protected].
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/opencog/CACfndaWPvtxJm-OEE%3DqJsSpAj85AvS4wO3qZ7CPn2ZfPuUkb%3Dw%40mail.gmail.com
> <https://groups.google.com/d/msgid/opencog/CACfndaWPvtxJm-OEE%3DqJsSpAj85AvS4wO3qZ7CPn2ZfPuUkb%3Dw%40mail.gmail.com?utm_medium=email&utm_source=footer>
> .
>


-- 
cassette tapes - analog TV - film cameras - you

-- 
You received this message because you are subscribed to the Google Groups 
"opencog" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To view this discussion on the web visit 
https://groups.google.com/d/msgid/opencog/CAHrUA34YMEr_S%2BFfovwcA2Gu6EQxgodhXrqmmJgxQ%2Bep%3D5jQxQ%40mail.gmail.com.

Reply via email to