Re: Walls and Communication

[glen e. p. ropella]

> In my Bee simulation Walls do two things: Occupy space and reduce the
> heatflow through the space they occupy. The simulation I am doing
> investigates if a cluster of bees can regulate its temperature when
> each bee bases its behaviour only on the information it gets about its
> local temperature. In this way, a bee will only know a wall exists if
> it bumps into it, although it will sense different heat sensations due
> to the wall.

Ahh, so the walls are big enough to contain more than one bee!
Can the clusters of bees change members?  And do you have more
than one cluster in your sims?  Final question (then I'll leave
you alone, I promise [grin]), do the walls move or change shape?

One of the issues we hope Swarm will help us deal with is the
preservation of the properties of swarms in spite of the turn
over in membership.  And if you've already got members of a
swarm surrounded by a membrane that's permeable to those members,
then you've got half of it in place.  A bee would move out of 
a cluster/swarm, through the membrane into space (where it would
either be a singleton member of the owner swarm or some special
transient swarm that contains only an individual bee), and then,
possibly, move into another swarm.

I'm not suggesting, of course, that you do this. [grin]  I'm just
throwing out one of the applications for possible heirarchical Swarm
models.

> With the emergence of language, I suppose, you could do something
> similar. It is only the bees that transmit heat with an ambient
> temperature outside the hive. A language model would involve the
> transmision of messages whcih may somehow be reduced in signal
> strength to a greater or lesser degree by objects(walls) in their
> path?

Maybe.  I don't know whether I want to put the transducers inside
the individual's boundary or on the boundary.  Right now, I imagine
putting the transducers at the boundary.  This seems most analogous
to animals.  (Even in the case of speech, where one might say that
the transducer is pretty deep inside the body, the transduction of
thought into sound waves doesn't *finish* until it leaves the 
mouth cavity.)

But, this will definitely be different from your Wall in that 
it will only surround one agent and it probably won't be passive.
Hmm...It sure would make things easier if it were passive, though,
wouldn't it?  All you do is decrease the amount of heat that
passes through the wall?  It works the same in both directions,
right?

> As for gaining a sense of self my bees seem more inclined to bunch
> together and form the traditional superorganism.......

Well, [grin] I'm  not really intending to generate a "cognitive"
sense of self, just an operational one.  I want the agents to be
able to "refer" to themselves and other objects.  All this need
amount to is a test on inclusion in a boundary (for "myself") and
distinction between two boundaries for "others."  I havn't talked
about object distinction, yet.  On the one hand, it's an easier
problem because it's been done already in well-understood ways.
(e.g. Motion detection, shading, etc.).  But, on the other hand,
I'd like the distinction of other objects to be done in the same
way one distinguishes oneself from other objects, which I'm not
sure has been done before.  And, of course, it's possible that
the agents don't need object distinction in the beginning.  Given
self distinction, object distinction could be built upon the 
atoms: myself, not-myself, and some message-conflict resolution
rule.

For instance, two agents acquire the ability to refer to their selves.
They then enter an encounter where one agent says, "Myself!"  Then
the other agent responds, "Myself!"  Both agents should then see the
same word for two different things.  I.e. "myself" refers to myself
and not-myself.  Then the conflict rule would kick in and tell
the agents to continue the conversation using randomly generated
words to refer to the other guy.  (I suppose one could make it so
that the randomly generated word could be used for "myself."  That
would result in agents naming themselves instead of using "myself.")

Hence, the only pre-programmed assumption would be that of uniqueness,
or that two different things cannot go by the same name.  Don't ask
me how to get classes of objects.  [grin]

glen