Re: Anticipatory Systems

[Jerry Zhu <***>]

Howard and Tim,

robots are programmed to learn from "experience"
supervised or not.  Artificial neural networks are
just mathematic algorithms. They just do what they are
told by deductively representing the model of
interacion within context,  there is no new
information or knowledge created.  If the definition
of AS is that "future change affect current change"
then a machine can be anticipatory, it is algorithmic
based.

Living systems create hypothesis from observation and
existing theories.  These hypothesises are new
information to explain new observation and accordingly
create new models.  

Algorithmic-based never transits to model-based for
ASs if them are both defined as ASs. We never create
life artifically.  Or machines we create can never
transit into life.

I think whether AS being complex or simple systems is
naming issue.  I saw some papers that classify AS as
having two kinds: algorithmic based and model based as
I posted.

Jerry




--- Howard Pattee <***> wrote:

> Tim,
> 
> Many robots are controlled by neural nets that are
> not programmed but that 
> learn from experience, e.g., see
> http://www.evobot.com/
> 
> I think there is going to be a more or less
> continuous transition from 
> simulated anticipatory systems to real
> (indistinguishable from living) 
> anticipatory systems.
> 
> Howard
> 
> At 11:32 AM 2/3/05 -0500, you wrote:
> >I would disagree that digital robots are qualified
> to be called
> >"anticipatory systems" as Rosen used the term.
> Rather, I would say that they
> >are simulations of anticipatory systems.
> Anticipatory systems are complex
> >systems [AS 425, EL 199]. A digital robot simulates
> in software the
> >interaction of the components (model, effectors,
> etc.) that would be
> >implemented in 'hardware' in a true anticipatory
> system. An anticipatory
> >system embodies impredicative relations among these
> components.
> >
> >I would agree that a digital robot can demonstrate
> an "anticipatory mode of
> >control". Judith some time ago pointed out that a
> camera diaphragm mechanism
> >[AS 41-42] also demonstrates an "anticipatory mode
> of control" but that it
> >is not thereby an anticipatory *system*.
> >
> >Regards,
> >Tim
> >
> >
> > > -----Original Message-----
> > > From: ROSEN Forum
> [mailto:*** Behalf Of Jerry
> > > Zhu
> > > Sent: Thursday, February 03, 2005 10:40 AM
> > > To: ***
> > > Subject: Re: Anticipatory Systems
> > >
> > >
> > > Judith,
> > >
> > > Both feedforward and feedback control are common
> in
> > > machine based control system. They are routined
> taught
> > > in undergraduate texts.
> > >
> > > Anticipatory systems can be simple systems being
> > > algorithmic based such as electrical or
> mechanical
> > > systems.
> > >
> > > Biological systems (1st and 2nd order
> autopoiesis) are
> > > anticipatory and model-based.  They are
> subcategory of
> > > living systems.  Social systems (3rd and 4th
> order)
> > > are different kind model-based anticipatory
> systems.
> > >
> > > Model based ASs have inductive capacity that
> > > characterise learning systems.  The implication
> is
> > > that computers are not learning machines.
> > >
> > > Jerry
> > >
> > > --- Judith Rosen <***>
> wrote:
> > >
> > > > Hi Jerry,
> > > >
> > > > One of the applications for studying natural
> > > > anticipatory systems is
> > > > to develop the ability to create anticipatory
> > > > controls for our own
> > > > technological needs. He used a high-tech
> camera as
> > > > an example of a
> > > > machine with anticipation built in. He
> recommended
> > > > that our modes of
> > > > government and city planning, social system
> > > > analysis, etc, all be
> > > > examined for ways to build in a "feed-forward"
> > > > control system rather
> > > > than purely feed-back. It's important to note
> that
> > > > while all living
> > > > systems are anticipatory they also have the
> > > > capability for purely
> > > > reactive behavior as well. That's something
> all
> > > > systems possess. But
> > > > anticipatory controls are superior under most
> > > > circumstances because
> > > > they prevent the system from entering an
> "error
> > > > state". An ounce of
> > > > prevention is worth a pound of cure!
> > > >
> > > > I would dispute that these encoded models are
> all in
> > > > the genes, Jerry.
> > > > There's so much more that's going on than just
> > > > genetics.
> > > >
> > > > Judith
> > > >
> > > >
> > > >   ----- Original Message -----
> > > >   From: Jerry Zhu
> > > >   To: ***
> > > >   Sent: Thursday, February 03, 2005 9:54 AM
> > > >   Subject: [ROSEN] Anticipatory Systems
> > > >
> > > >
> > > >   --- Judith Rosen <***>
> > > > wrote:
> > > >
> > > >   How the information in these internal
> predictive
> > > >   models is encoded, how it is integrated with
> > > > "real"
> > > >   behavior of self/environment, how it acts on
> > > > system
> > > >   behavior... these are all wide open areas
> which
> > > > are
> > > >   just screaming for one of you guys to
> pursue.
> > > >
> > > >   Judith,
> > > >
> > > >   A robot is also qaulified as being
> anticipatory.
> > > > It is
> > > >   quit common practice to have machine
> learning,
> > > > pattern
> > > >   recognition to learn environment and
> anticipate
> > > >   consequances of future actions so to adjust
> > > > current
> > > >   action. A robot can learn environment and
> get
> > > > around
> > > >   obstacles along the way.  These type of ASs
> are
> > > >   algorithmic such as computers and they are
> closed
> > > > in
> > > >   the sense that the change of environment is
> > > >   predetermined.  The description of self and
> > > >   environment is syntactic. Its mechanism is
> > > > deduction.
> > > >
> > > >   Life as being anticipatory is model-based
> intead
> > > > of
> > > >   algorithms based.  A insect changes its
> behavior
> > > > to
> > > >   adapt to evironmental change not thro
> genotype but
> > > >   thro mutation of the genes.  New genes are
> created
> > > > in
> > > >   next generation that are hypothesis not in
> older
> > > >   generations.  New behaviors are generated by
> > > >   interpretate the new genes.  This process is
> > > > inductive
> > > >   and semantic.  New knowledge is created by
> > > > enlarging
> > > >   the understanding of environment and self
> thro
> > > > these
> > > >   new genes or hypothesis.
> > > >
> > > >   Jerry
> > > >
> > > >
> > > >
> > > >  
> __________________________________________________
> > > >   Do You Yahoo!?
> 
=== message truncated ===


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com