Alt-BEAM Archive
Message #00575
To: "'M le Roux'" mleroux@itec.co.za, beam@corp.sgi.com
From: Wilf Rigter Wilf.Rigter@powertech.bc.ca
Date: Wed, 17 Feb 1999 23:49:56 -0800
Subject: [alt-beam] BEAM - TECHNOLOGY OF EVOLUTION
Hello Juan,
You bring up an important point. I think that BEAM robot "survivability" has
been consciously isolated from BEAM "applications". Read Living Machines
and then make up your own mind.
The following are my own opinions and based in part on my reading of the
literature and the opinions of others. They are also an extrapolation of
the more conventional explanations of BEAM technology.
The Technology of
Evolution
wilf rigter
(c) 1999
In the technology of evolution we focus our attention on the fundamental
problems of emulating biological evolution with technological evolution.
The technology that follows bears a resemblance to biological systems but is
not a "copy" of biological systems. The emphasis on evolution in
development of BEAM technology, is sharply different from other design
philosophies which views legacy technologies only in terms of "problems of
backward compatibility" to be phased out as soon as possible. This reduces
the range of evolutionary pathways through a process of fossilization of
technological history both in the technology itself and the minds of it's
creators.
Take the case designing a robot walker with a electromechanical system based
on biological legs.
There are two distinctly different approaches to this problem.
1. Reverse engineering
2. Evolutionary design
Reverse engineering decomposes the form and function of "biological legs"
into subsystems (skeleton, muscle, circulation and nervous system) and
further down into "components" (bone, connective tissue, blood, neurons )
and "elements" (ATP, ion channels) but I'm regressing . Next equivalent
electronic/mechanical components and subsystems are found and are combined
into a walker design which is then optimized to resemble "biological legs"
as close as possible.
It may be said that in this engineering philosophy form is derived from it's
function (ie application).
In biological systems, the evolution of legs can be traced as a series of
adaptations starting with mobile multicellular organisms in the primordial
oceans increasing their access to food/energy through mobility to Ben
Johnson increasing his access to money/energy in the 100 yd dash. Despite
specialization of human legs for walking they retain other secondary
functions such as climbing, swimming, kicking, crushing, curling up and
picking up pencils from the floor. Legs in other species are specialized in
different ways like running, climbing, paddling etc but also have a range of
secondary functions. In general, despite apparent specialization,
"biological legs" represent and retain the memory of their evolutionary
process. Biological "legs" are remarkably versatile and adaptive to the
current environments and retain the ability to evolve into new niches.
The evolutionary design approach is acutely aware that biological systems
are not "designed" but are instead a snap shot of an ongoing adaptive
process common to all life.
Therefore the evolutionary machine design duplicates natural evolution in an
iterative process of inventing primitive electro/mechanical elements and
subsystems whose adaptation to the environment and reproductive success is
based not on sex but on survival of the fittest technology (may the best
tech win). The most successful of these elements and subsystems are then
successively combined into higher and more complex systems culminating in
"useful" applications similar to species evolved in to ecological niches.
One could argue that in evolution "function follows form" since adaptation
is largely a series of mutations of form with emergent functions that favor
survival.
As a result the current batch of "walkers" have "legs" and mimic insectoid
forms in their mode of locomotion. But the basic BEAM "walker" mechanisms
are discoveries as much as designs and are quite unique and potentially
broad in their applications.
This is the exciting characteristic of the evolutionary design approach: we
don't know exactly what we are designing. It's much more a process of
discovery and invention rather than engineering.
So BEAM technology as a form of evolutionary design is barely off the
starting block in terms of it's own evolution, however within a few thousand
generations (about 10 yrs with a focused group effort of people building
and trying out designs) BEAM's "sum over histories" phase space will have
increased by several orders magnitudes and I would expect the fundamental
forms and functions of advanced BEAM organisms to have much more in common
with the self-similarity and adaptive mechanisms of living biological
organism than with computers. Then BEAM applications can be "filled" by
nudging a branch of BEAM ecology into the "application niche". At that
time BEAM can truly lay claim to be the technology of LIVING MACHINES.
Although BEAM function follows form and we would like to see our BEAM
creations reach their full evolutionary potential, some ethical constraints
must be imposed on our own involvement in this process based on the Asimov's
laws of robotics. Despite the importance of "survival" at this stage in
BEAM evolution we should note that our own survival as a species comes first
("zeroth law") and that survival of the robot is the subservient third law.
If we are successful at facilitating the evolution of "living machines" we
may also have created potential competitors for our own species! So I say
to you, "Observe Asimov's Laws and please do not permit the evolution of
self-replicating carnivorous living machines especially if they are great
survivors"
Wilf Rigter mailto:wilf.rigter@powertech.bc.ca
> -----Original Message-----
> From: M le Roux [SMTP:mleroux@itec.co.za]
> Sent: Monday, February 15, 1999 3:20 PM
> To: beam@corp.sgi.com
> Subject: Re: hmmm...i wonder which subject
>
> >Having a bot take on a job is a little over the current state of BEAM
> >robotics, and is not really keeping in tune with the general "bottom up"
> >approach..We aren't trying to make a bot that can get from point A to
> >point B so that it can accomplish a set goal. The BEAM/bottom up
> >approach is to make a bot that can start from point A and *not get
> >itself killed* without even trying to get to point B..
>
>
> How will we know when we have reached this objective?
> Surely the required level of surviveability depends on the intended
> application?
> I agree that one of the fundamental aspects of BEAM is
> surviveability, but I don't think is should be an isolated (or
> exclusive)goal. We will only be limiting ourselves.
>
> It should rather be seen as an integral (and important) part of a larger
> and
> more far-sighted design process. (not a hurdle to be overcome before any
> further development can take place)
>
> Being new to BEAM and the list, I hope I'm not being too presumptuous :)
>
> Juan
>
>
>
>
>
------------------------------------------------------------------------
eGroup home: http://www.eGroups.com/list/alt-beam
Free Web-based e-mail groups by eGroups.com
Home