Alt-BEAM Archive

Message #05241



To: beam@sgiblab.sgi.com
From: Steven Bolt sbolt@xs4all.nl
Date: Sat, 10 Jul 1999 22:08:01 +0200 (CEST)
Subject: [alt-beam] Re: Chaotic controllers (was Something funny with the 1382)


On Sat, 10 Jul 1999, Bob Shannon wrote:

> I'm on it, but while I'm getting the stuff together for the contest, feel
> free to ask away.

I'll await your posting. Otherwise I won't know what questions to
ask :)

> > There is a strong and deterministic connection between the
> > environment and behavior of animals and even humans. The usual
> > example is the ant - a relatively simple piece of biological
> > machinery, which follows something like a dozen fixed rules. Put it
> > in a simple environment - like a Robot Jurassic Parc - and what it
> > does looks no more intelligent, capable or complex than the
> > behaviour of today's robots.
>
> I disagree.
>
> The obstacle aviodance of an ant are vastly bettern than any BEAM
> robot. Its ability to handel terrain is also vastly better, as
> its control system and drive system are well matched in ability,
> this is not true for beam at all.
> Your underestimating the ant here, bigtime. The ants sensors and
> processing power are far beyond any BEAM design.

Um, that's the point I tried to make - here and earlier. The ant is
so much better connected to its complex, real world environment.
That connection, the bandwith of its sensor system we might say, is
what allows that environment to turn a fairly mindless biochemical
robot into a creature whose actions have purpose and meaning.

*But*, the great advantage of the ant's sensors evaporates almost
completely when it is locked in to the greatly simplified
environment offered by a Robot Jurassic Parc. From this I conclude
that the long term strategy for small robot development ought to
aim at better, cheaper, simpler, smarter, longer range sensors.

In the mean time, if you want current robot technology to do
something interesting, you will usually have to engineer the
environment to match the present sensor limitations - like was
done with the Egg Hunt Playground. Which as a contest puts a high
value on long range Egg sensors. An idea for a simple scanning
laser device is shown on

http://www.xs4all.nl/~sbolt/Other/e-EggHuntDiscussion_090799.htm

> After all, what do you think its doing with thousands of times
> more (real) neurons than any BEAM design will ever have?

Imho it's a matter of connectivity to the environment, of sensors
rather than brains. And before we revisit a previous discussion,
yes, those sensors do considerable but entirely automatic
processing. Our vision system presents us with a threedimensional,
color-corrected image of the world; we don't even have to
consciously recognize tables, chairs and faces in a crowd. To me,
the thinking brain is a layer on top of that.

And plenty of BEAMish thinking power can be done by the
tenthousands of transistors in a modern uC, right?

> > But look at what happens when you put that ant in the `real'
> > world! All of a sudden its actions acquire purpose and meaning.
> > Now where did that come from?
>
> I disagree with this observation. The ants purpose and meaning
> do not come from its environment. Ants colonise forign
> environments after all, they adapt.

There seems to be a pretty wide consensus about the complexity of
ant behaviour being caused by the interaction with its environment.
Simple environments reduce the ant's behaviour. When ants colonize
environments in which the majority wouldn't fare to well, natural
variation comes into play. The ants which accidentally fit the new
environment well enough move in. If the circumstances are right,
speciation may follow, and after a relatively short while, you have
a new species of ant.

> > It's not supplied by the ant's brain. Purpose and meaning happen
> > because the ant and its environment evolved together, into a single
> > complex entity, which doesn't work when you take it apart.
>
> Clearly it is supplied by the ants brain! Thats why the brain
> evolved Steven! Your being fooled by looking at a single ant.
> The ant evolved as a colony and must be evaluated this same way.

I looked at the individual ant to evaluate its behaviour in the
complex real world as opposed to simplified RJP environments. But
the argument doesn't change when you look at an ant colony as a
single, multibodied creature. That creature also needs its complex
environment to give its behaviour meaning. In an RJP, it would
merely die.

> > > We will not design, nor evolve higher leveled controllers this
> > > way. Biological neurons get around this problem by having the
> > > ability to change their dendritic connections dynamically,
> > > something Nv's can never do.
> >
> > Emulated Nv's can do just that, if I understood Terry's postings
> > correctly.
>
> The parts of Terry's code that shuffle the interconnections is
> not part of the Nv emulation (Bicore emulation technically) in
> any way. Its a totally seperate functional block that is quite
> alien to BEAM tech. Its a common enough approach in the CPU
> world of robotics however.

That's mere flag waving. You get upset if something isn't on Mark
T's list and I call it BEAM anyway. Well, it pleases me to do so.
Terry's shuffling of the connections between emulated Nervous
thingies is a nice advance of BEAM tech in my book. Chapeau to him,
and to previous BEAM in so far as it inspired him.

Best,

Steve

----------------------------------------------------------------------
# sbolt@xs4all.nl # Steven Bolt # popular science monthly KIJK #
----------------------------------------------------------------------




------------------------------------------------------------------------

eGroups.com home: http://www.egroups.com/group/alt-beam
http://www.egroups.com - Simplifying group communications



Home