Skip to Content

Background Information - Data Ecologies 2003


Simulations of physics, Physics as simulation.

A two day workshop investigating the interplay between the simulation of physics for the creation of artificial (data) ecologies, and the possibility that physics (and thus, all ecosystems) is itself, a data process.

This workshop evolved from the title: a small seed that sprouted, bloomed, mutated and turned, with all the necessary branches of a growth process, into something completely different.

Perhaps the simplest: "workshop." Thus the days should be most simply used to work over the ideas and implications that the other two words, in their reinforcing reflexive combinatorics, lead to. The ideas and their implications should be paid attention to, the free exchange of ideas among an audience of interested participants will be encouraged.

"Data" leads us towards the realms of the bit, the abstraction, that which is distinct from physicality, order or disorder, complexity, algorithmics and other textualities.

"Ecology" gets us into the realm of sloppy goop, the primal soup, structures of highly physical, biological processes that reiterate, the regenerate, breed, eat, reuse and generally get as far from the cleanliness of data as possible.

Of course, getting these two things together leads us to the ecology of data (the movement of data in networks? That's not what we thought that we meant), or the data of ecology (logs of plankton travel? Hmm, neither). As the terms begin to intermingle, we get artificial evolution, data niches, simulation of biological systems, genetic codes as data collections, genetic manipulations of data collections...: the autocatalyticlist seems not to want to stop!

In order to escape this Sorcerer's Apprentice situation, we skimmed the top of the brew and removed a small fragment with two flipsides, both of which seem to be of relevance: ecosystems in simulated physics and the question of physics being a simulation.
Evolution in Simulated Physics

We cannot help ourselves: pure data is not our thing. Time's Up hasspent too much time running around building things to be satisfiedwith the purely informational. If we are to delve into simulated evolution, then we need to get our fingers dirty and to know that there is a certain aspect of physical "realism" behind the plans.

In the a-life heydays of the early 90s Tom Ray's work showed that an ecosystem of sorts, with the spontaneous evolution of all sorts of niches and parasitisms, could exist on a purely informational level. These results, surprising and full of motivations for further work, have led to all sorts of evolutionary simulations and systems. The presentation of these results has often run up against the barriers of understanding: making the processes visible to an interested but not fully comprehensive viewer is difficult. Since then, other work has extended towards looking at things that make more intuitive sense for the viewer of such artificial systems. Karl Sim's work on evolving critters to actually _do_ something, in this case to walk, to swim, to act in the simplified world that he created, is part of this sequence. Of course, this moves away from ecosystems and into the realm of virtual husbandry, but the motivation to be able to see the artificial evolution of action and motion was important.

The development of more and more "realistic" physical models has led to an increase in the usefulness of the results. The ability to not just evolve shapes as per Dawkin's early experiments, not just bodies with the evolution of L-systems, not just controller systems as per the swathe of evolutionary robotics, rather the complete coevolution of "bodies and brains" has become the focus of interesting work. One end result of this work, the construction of actual machines based upon the results of evolutionary development, has led to some extremely intriguing machines and methods of movement.

As the construction of physical systems is a little more than trivial, this end of development has been placed in the "interesting but too hard" basket. The workshop aspect will instead keep its feet at least vaguely in contact with the ground and will look into the more developed of the available systems for the evolution of bodies and their control in a simulated physics. In particular, we will focus upon the system Framsticks developed at the Poznan University of Technology in Poland. This system, developed since 1997, allows several levels of evolutionary force to act upon bodies and their controlling systems. The system has been greeted with a broad range of uses. The workshop will introduce the system and outline some of the ways in which it differs from other such systems, where it has been successfully used, what can be done with it. Then a more intense introduction to the system should enable the workshop participants to get into the use and misuse of the system for their own purposes.

In particular the ability of the Framsticks system to maintain some sort of semi/pseudo physical reality yet run complex processes of spontaneous, i.e. non goal-oriented evolution, is of primary interest. The interfacing of other data systems to an artificial pseudo-physical is relevant and of interest here.
Physics is a Simulation?

The flipside theme for the workshop, taking the second day, is the following question:

What if the world is actually, at its finest level, computational?

That is, what if the idea of simulating physics as we understand it (a continuous process) is somehow redundant because physics is in fact an informational process.

This radical proposal, often termed the "Zuse hypothesis" after its earliest proposer, Konrad Zuse, has recently gained some serious scientific interest. Ed Fredkin's version of the thesis, sometimes argued with, sometimes argued strongly for by none other than Richard Feynmann, has gained more attention recently. A popularisation of the idea by Stephen Wolfram in his recent self-published tome has raised some other forms of attention.

The thesis leads to some strange conclusions. Some of these begin to differ from what is commonly accepted as the standard model of physics. Looking into some of these ideas, for completely different reasons, Hartwig Thim has conducted experiments that begin to disagree with the standard model in the same sort of way that Zuse's thesis would disagree with it.

Although this meeting is not meant to become some sort of meeting of arcane physicists, it is planned that the ideas that are collected in this model should receive quite some critical and other attention. Although most of us are deeply involved in computer-based activity, we are happy that there is in fact a "real world" out there. This group propose that the equations of our world are perhaps more relevant than the objects whose motion they are meant to model, that the terms real virtuality and virtual rality are, perhaps, interchangable.

URLs:

http://www.frams.poznan.pl/

http://www.spiderland.org/breve/

http://computing.tay.ac.uk/timtaylor/cobb/

http://web.genarts.com/karl/evolved-virtual-creatures.html

http://www.numa.uni-linz.ac.at/Board/thim.html

http://fodok.uni-linz.ac.at/fodok/publikation.xsql?PUB_ID=11801

http://tph.tuwien.ac.at/~svozil/

http://www.bottomlayer.com/

http://www.digitalphilosophy.org/

http://www.idsia.ch/~juergen/digitalphysics.html