View of Arrows from a College Nobody

[Alexander Clouter]

Firstly, I haven't followed a Computer Science university course and so
am not going to 'pretend' to know anything (or even what I am talking
about :-) ) but I did decide to churn through Brians Arrows paper to see
what I could learn.  I joined this group for the simple reason that I
wanted to learn more, especially unique methods to tackle problems
(information handling in my case) by seeing how others went about
solving problems.  OS design for me seemed to be the way ahead, plus I
always liked the idea of having my machine running on my code from the
bottom up.  The only way I've found to learn things is to be thrown in
the deep end but without armbands, hence why I joined the Tunes mailing
list.

Anyway...reading through Brians paper I found it enlightening, I didn't
understand much of it, but what I did I think I gained a lot from it.  I
always had looked at information as something really to be entered into
a system, tweaked stored and recalled.  I always imagined a system that
attempted, no matter how primitive, to understand the data to some
degree and thus relate it to other information; but I always thought of
this as a far too a complicated project to do, anyone fancy building a
brain??
	Reading Brians paper I think he has found a way to approach this
concept, if not at least something to build upon such as the
interpretation of Arrows having colour by June Kerby.  However I cannot
see how this could be implemented for an OS, maybe as a well integrated
Expert System with several domains depending upon the task at hand.  It
seems the Arrow system constructs a 'algebraic' neural net (NN) for the
information to dwell in that we can all contemplate and understand whats
going on, maybe not fully, and how the system
works/processes/understands that information.  As information is added
to the Arrow system it becomes integrated into the whole domain, while
the system can process the information, whilst it is inside the domain I
cannot see how it could be extracted from one domain with any
conclusions about that information and be placed in another!  This is
probably where I don't understand the Arrow system but no matter how I
tried I could figure a method for this to occur.  The only way would be
to take the whole domain (imagine how big that would be) and move that,
defeating the object as how would you go about expanding a domain with a
totally different domain?  I relate the problem of once you train a NN
with 'rules', throw away those rules and now try to extract the 'rules'
from your NN.  While you can draw conclusions from it, moving 'rules'
from one NN to another is short from impossible I would imagine!

	This is the only thing I find awkward, I probably am completely wrong
but its just how I see it.  I do intend to re-read Brians paper once I
get a hard copy done, several times mind you, so you may find a
"Withdraw previous message" in the near future ;-)  The whole point in
an OS is being able to spread information between domains (different
machines) as transparently as possible.  With Arrows I cannot see how
this can be done.  One way round it though would be a Parallel System of
networks working on one domain, each computer would have a user at it
concluding and entering information into the domain whilst the whole
network could grow and everyone would have access to the information and
its conclusions.

Well thats enough from me, I hope my grammar and terminology isn't too
bad but I do expect some things to of slipped in, mainly as I haven't
just read what I have written ;-)

Alex

P.S.  Just my $.02

-- 
 **       ((__))   Alexander "Drutt" Clouter 
  \\      ((oo))      e-mail: drutt @ bayesian . prestel . co . uk
   \\------\\//                (remove those spaces to e-mail me)
    ||      ||        WWW   : http://www2.prestel.co.uk/bayesian/alex
    |||----|||                        (under construction)
    ~~~    ~~~        equip : Acorn A440 with ARM3, 4Mb and 500Mb HD
   Cow during an              BBC B 32k, 2x5.25 drive with DDFS
    Earthquake                P-(yuk)-II 266Mhz, stable surprisingly