[Date Prev][Date Next][Thread Prev][Thread Next]   [Date Index] [Thread Index] [Author Index] 

Re: machines vs. living systems

Tim Gwinn wrote: 
TG: Have you read Life Itself yet? That would help answer many of your
questions below.

Yes. That and "Fundamentals...". I have skimmed Anticipatory Systems and I have Essays on Life Itself, but haven't cracked it yet.


A realistic enough model of this system will not be simulable.
> 
TG: The set of models for a system are dependant upon how the system is
defined. Define it to include aspects like atomic interactions and you can
wind up with a complex system. Define it in coarser terms like pistons,
crankshafts, etc., then it is a simple system. You must be precise in your
system definition. It does not suffice to talk about 'practical'
differences - that is too vague.

OK. Before we go too far down this path, I want to back up a bit. Since this list is about RR's work, it's appropriate to dive right into the rich universe he created with his stories. But, for a collection of stories to be useful, they have to map to the real world both in technique (tools and methods) and in understanding.

So, when I talk about a naturally occuring system (like a storm) or an ontologically extant object (like a car engine), I'm talking about something we don't _define_. We may _build_ it. But we don't _define_ it. An engine is not a logical (3rd world, platonic) object. It's a real thing that sits in my driveway.

When RR states: "But this peculiarity stems only from my expression of these concepts in terms of the models of N, rather than try to talk directly about N itself." pg 203 of LI (Columbia)

He seems to be referring directly to the fact that naturally occuring systems aren't defined. And that's why, when we engage in inferential processes, we restrict ourselves to talking about what we can define, like formal systems.

Now, when I talk about the engine that sits in my driveway, I have the option of defining a formal system for it that is very fine-grained and rich in it's expressiveness (and therefore more susceptible to both complicatedness and complexity) or defining a formal system for it that is very coarse-grained (and therefore less susceptible to complicatedness and complexity). (Note that I allow Judith's point that the coarse-grained and the fine-grained can be either complicated but not complex or complex but not complicated... so, I'm not suggesting that the complexity is a function of granularity... only that it can be correlated.)

Given that, there are situations that will obtain with the engine that sits in my driveway that will require a fine-grained formal system to describe. If the formal system is too coarse, then it will not describe, for example, "pinging" (which I think is a symptom associated with the flammable wave-front when the gasoline is ignited in more than one spot in the cylinder -- though I could be wrong about that [grin].)

So, in order for a formal system to be expressive enough to capture _some_ phenomena, it has to reach a certain level of fine granularity. And my suggestion a couple of rounds back in the conversation was that _any_ formal system that is rich enough to express _both_ the coarse-grained phenomena (like ideal combustion, ideal camshafts, etc) _and_ fine-grained phenomena like "pinging" will be a _complex_ formal system because it will need to describe complex physical phenomena.

If the formal system isn't capable of complex (meaning plaited) theorems, then it is incapable of representing the complex (meaning plaited) phenomena in the actual engine that sits in my driveway.

So, what all the above rambling is about is not _just_ the Rosennean focus on itra- and inter-formalism gymnastics; but, it's about practical requirements for formal systems in engineering and even science to some extent. (The thread was started by suggesting that the fabrication of an organism by putting genes together like Legos, after all. So, what I'd like clarification from Rosenneans out there has to do, fundamentally, with how we map all the platonic gymnastics in this mathematical/logical world to the world we experience directly with our sensors and effectors.)

For Rosen's more formal use of the term "machine", it makes me think
that "machines" are platonic and will only help in the world of
formalisms and inference.


TG: 'Machines' are not discovered in the world, they are defined by us, and
we then realize them from these formal descriptions. We build computers to
conform to formal definitions, such as Turing-computability.

This goes right to the heart of what I'm talking about. So, pleaseplease don't think I'm being trite with the following comment. I'm not.

"Machines" _are_ discovered in the world. This is part of the whole "stigmergy" and "emergent phenomena" stories, including the "surprise" (modeling and estimation) story Dan referred to earlier. There are whole generations of humans that grow up in the "inner city"... Their "naturally occuring systems" consist of buildings, glass, steel, concrete, planes, trains, and automobiles. The people that grow up living on top of and inside these artifacts don't experience them the same way that the inventors who created them do.

(Granted, if we're predisposed as children to break things and look inside, or if we go to college and learn anthropology and mathematics, then we begin to invest ourselves in the stories that rely on the distinction between artifice and nature. But, imagine the _large_ percentage of us who don't tinker and don't go to college.)

As another example, witness the "script kiddies" that create most of the virii and find most of the exploits for machines on the internet. There are transcripts of their IRC traffic. Read those transcripts. They don't know or care, for the most part, _why_ sendmail was created and when or how it compares to other MTAs like Exim. All they know and care about is that sendmail has been around since before they were born and, therefore, is pretty rock solid. So, if they find a way to overrun a buffer in sendmail, they will be rock stars amongst their peers.

Sendmail, as a piece of software is just about as "formal" as you can get, though it does live inside a larger system that is not really a formal system because it's an ill-defined collection of real devices that can behave outside the defined rules of inference.

The reason I claim this is important and appropriate for this email list is because there are _artifacts_ (things we engineered) that are placed right in front of us and we cannot tell whether they are alive or not.

E.g. take, on one hand, the internet, including all its gory details of script kiddies, predatory solicitation, IP theft, etc. and, on the other hand, a human. Both are embedded in and fundamentally dependent on a larger system (be it a city or a forest). For those of you who will immediately say "The internet is not alive; it is a machine", what do you see that I do not see?

The term loses meaning when applied to the
dirty world of reality.


TG: The formal system 'machine' and the material realization of it are not
the same system. You are trying to equate the two while simultaneously
applying a different system definition to the material system, so of course
they will then differ.

No. I'm not equating the two. But, I am asking for how the inference methods RR developed will help me use the formal systems called "machines" to study extant objects.

Because the term 'machine' is used in a very specific and narrow
sense here, this may clarify the distinction made here between
machines and living systems. If not, just ask some more questions. :)


I wouldn't want to separate the specific sense Rosen uses from the
vernacular sense... or at least I wouldn't want to overly separate them.


TG: You have to. When the line is not clearly drawn is when confusion sets
in.

But, if you draw the line too clearly, then the _usefulness_ of the formal methods disappears. The thread started with the discussion of a Frankensteinian construction of new monster from scratch. If we draw the line separating the formal methods from practical engineering and construction, then we turn ourselves into pure philosophers who can and should be ignored by people like Dr. Venter.

I'm not aware of two formulations.  I'm referring to "Theory of
Self-Reproducing Automata".  What was the other?


TG: There is a kinematic automaton version and a cellular automaton version.

Ahhh. I'm referring to the CA automaton. To my knowledge, the others weren't developed to anywhere near the extent the CA was. But, I'd love to see more info on the other 4. (It sucks that vN's writings aren't more readily available... if only we had a champion like Judith to help preserve vN's work.)

In any case, my only reason for bringing up vN was to suggest that he made significant strides in making "machines" look more like "organisms". If you imagine a Venn diagram with two circles, one being the set of "machines" and the other being the set of "organisms". It's safe to say that some things (like a pair of scissors) fall squarely under "machines" and other things (like a human) fall squarely under "organisms". Somewhere in there could be a 3rd circle "things that reproduce". I would suggest that vN moved that 3rd circle so that it intersected with "machines".

Anyway, I'm coming to the conclusion that this discussion doesn't really belong on this list. Perhaps I should take it to a forum more focussed on engineering and constructive modeling and less focussed on inferential methods? As list demiurge, what do you think?

--
glen e. p. ropella              =><=                Hail Eris!
H: 503-630-4505                       http://ropella.net/~gepr
M: 503-971-3846                        http://tempusdictum.com