Re: Computational complexity

[Tim Gwinn <***>]

Hi JohnK,

Thanks for the link.

I know that these are only demos, but I simply do not find them particularly
interesting or novel. The old pseudo-AI program "Life" (I think it was
called) decades ago produced similar patterns.

In my estimation, "agent-based modelling" amounts to little more than an
extension of today's common object-oriented programming paradigm. Rather
than write huge monolithic programs or breaking code into main programs with
subroutine calls, object-oriented programming has programs "send messages"
to other code-objects. Because the objects are largely independent of the
main code or other objects, they can (in principle, anyway) improve code
stability and facilitate writing reusable code modules. Its just one of many
ways of writing code. ABM appears to follow similar distribution of code
notions, and instead of displaying on the screen only output from the main
code, what ABM displays on the screen are the outputs of instances of some
of the code objects.

I agree that just as coding a monolithic program into subroutines calls or
using an object-oriented coding will usually make coding and debugging alot
easier and reduce coding errors, there's nothing particularly novel about
it, nor is ABM particularly novel. It is a coding method that is useful, but
it does not break any fundamental barriers. All these are entirely simple
systems.

I disagree that the ABM "approach takes a rule set and derives unpredictable
behavior and patterns". The only thing that is unpredictable is that the
random number seed(s) that are used for each run of the program, and that
simply because it is hidden from the person running the program, so that
when the program runs we see the display change in ways we could not readily
predict unless we knew those seed numbers. In fact, with the same seeds,
each run would be identical and entirely predicable: the program itself can
introduce *no* fundamental novelty or alteration of output, aside from the
intentionally introduced variations via random-number seeds.

Since their novelty rests on RNG seeds and their effect on initial
conditions, ABMs may be instances of nonlinear dynamical (chaotic) systems
that are discrete rather than continuous, but they are not at all complex,
in my opinion.

Regards,
Tim

> -----Original Message-----
> From: ROSEN Forum [mailto:*** Behalf Of John
> Kineman
> Sent: Friday, April 16, 2004 12:21 PM
> To: ***
> Subject: Computational complexity
>
>
> Hi, The following link is to a Windows executable that installs SWARM
> demos. This is the sytem Chris Langton developed at SFI.
>
> ftp://ftp.swarm.org/pub/swarm/binaries/w32/swarmdemos-2.1.1.exe
>
> I thought the 'dynamicgraph' demo gives a good idea of what is meant by
> computational complexity in reference to the earlier discussion about
> SFI, Langton, and von Neumann's approach. Some of the other demos don't
> run on my machine and none are documented, so it wasn't clear what they
> were about.
>
> In the dynamicgraph simulation you see nodes moving by some ruleset and
> forming relationships. From these rules various patterns form. The
> patterns that form cannot be predicted by any calculation other than
> running the iterative process, like a fractal.
>
> I think the way to describe the difference between this kind of
> simulation of what a "living" system behaves "like", i.e., similar
> patterns, and what Rosen was saying is required to get at what it is
> really doing, are several:
> 1. in how the probabilities are handled. Here they are assumed to be
> random variations within some constraint - there is no "function" that
> biases them.
> 2. in the rule set. While the agent-based approach takes a rule set and
> derives unpredictable behavior and patterns from it, it is still
> deterministic. The Rosennean approach suggests (requires) that the
> organism is also capable of changing the rule set. This is a level of
> functional feedback that isn't captured, of if it were, would also have
> to be treated as a random change.
>
> JJK
>
> PS I tested this file in Windows XP and it installs OK. Not all the
> demos work properly, but the dynamicgraph does. It does not contain any
> virusus or worms. It may not run correctly on all systems, but it should
> be safe to install. It can be uninstalled from the Windows Control
> Panel, Add/Remove Programs menu.