[Top] [All Lists]

Re: [ontolog-forum] Axiomatic ontology

To: "[ontolog-forum]" <ontolog-forum@xxxxxxxxxxxxxxxx>
From: Ed Barkmeyer <edbark@xxxxxxxx>
Date: Mon, 04 Feb 2008 12:17:55 -0500
Message-id: <47A748C3.3010204@xxxxxxxx>
Randall R Schulz wrote:
> On Friday 01 February 2008 19:18, Francis McCabe wrote:
>> ...
>> Being deterministic is not the same as being predictable. Chaos
>> Theory/ Fractals tells us that; as does the weather.
> Chaotic behavior occurs only in continuous systems. Digital computers 
> are discrete, and that is not an incidental or unimportant difference.     (01)

But "discrete" digital systems can be programmed to respond to events 
occuring in the outside world to the extent that they trigger sensors to 
emit digital signals.  That world is continuous, as is the change in 
current and voltage in many sensors.  As a consequence, if you scan the 
same image twice, you don't necessarily get identical bit patterns from 
the sensors, and algorithms that operate on the resulting image, 
although they are completely deterministic, do not generate identical 
results.  In some cases, for example, the algorithm recognizes a feature 
in one scan and not in another.    (02)

> Digital systems don't exhibit the phenomena associated with nonlinear 
> dynamics. There is no "sensitive dependence on initial conditions," 
> no "strange attractors," etc.    (03)

In one of my earlier careers, I was a data communications expert.  In 
all real-time programming, and particularly communications, you get all 
kinds of event patterns, and it is very typical to find "race 
conditions" in deterministic code:  If X occurs before Y, the software 
does A; if X occurs after Y, the software does B.  So first, when the 
events occur nearly simultaneously, you get different results depending 
on the exact timing.  Second, it is not whether X occurs before Y but 
rather whether X is detected in the algorithm before Y that determines 
whether the result is A or B.    (04)

[In one memorable case, a manufacturing workstation controller wired 
into the factory broadband network had a simple 2-state detector (a 
microswitch) in a parts receiving station.  As soon as a tray of parts 
was fed into the receiving station, it tripped the detector and the 
workstation took action, and in some cases that action involved 
communication with a higher-level manufacturing management system.  When 
the Automatic Guided Vehicle delivering the materials initiated an 
unload, it informed the management system as well, unfortunately by a 
very early wireless technology at 1200 bits per second.  So when the 
workstation asked (at 10 Mbps) the management system about the parts 
that had just arrived, the management system was unaware that any had. 
Of course, the AGV unload had to happen before the workstation detected 
the insertion, but in the management system, the notifications of the 
two events occurred out of order, and the software took the wrong action 
(in effect, it said to the workstation: "there are no such parts", which 
caused serious problems with the intended job).]    (05)

The problem here is that people tend to think of digital computer 
systems as being a single processor with a single, algorithmically 
determined, "thread of control".  The reality is that all modern 
computational systems involve multiple communicating processors, even 
inside the box, and many interesting systems involve communication among 
multiple autonomous agents across boxes, intentionally and 
unintentionally.  (Unintentional communication occurs when some agent 
takes a physical action that is detected by another.)    (06)

And there is a related behavior that I think of as "the Heisenberg 
effect" -- attempting to acquire information from a system provides 
information and alters the system.  When you access a web page, the fact 
that you accessed it becomes part of the knowledge held (and possibly 
used) somewhere on the Web.    (07)

Now, algorithmic parsing of a given natural language text is perforce 
deterministic, and even when the algorithm is Bayesian and covers a 
corpus of multiple texts, it is still deterministic if the corpus is 
fixed or carefully controlled.  But if you allow the corpus to change 
dynamically, the results of the deterministic algorithms are not 
deterministic, and if reasoning algorithms are then employed, the 
results are chaotic -- repeating the same experiment 3 seconds later can 
produce contradictory results.  Think of trying to date a piece of 
ancient pottery while the Egyptologists are adding new timelines to the 
literature, revising some respected sources, and burning "outdated" books.    (08)

-Ed    (09)

Edward J. Barkmeyer                        Email: edbark@xxxxxxxx
National Institute of Standards & Technology
Manufacturing Systems Integration Division
100 Bureau Drive, Stop 8263                Tel: +1 301-975-3528
Gaithersburg, MD 20899-8263                FAX: +1 301-975-4694    (010)

"The opinions expressed above do not reflect consensus of NIST,
  and have not been reviewed by any Government authority."    (011)

Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/  
Subscribe/Config: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/  
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/ 
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx    (012)

<Prev in Thread] Current Thread [Next in Thread>