|
Symposium
Mark
Rudolph
Metaforms Methodology
Presented
Saturday May 11th at 2 pm
One
of the constraints on the flourishing of 3D content on the
web and hence on interactive 3D productions is tediousness
of modeling in 3D to produce sets and props. The need for
skilled hand work on each individual piece using expensive
tools drives up the cost of production and slows completion
times. The situation resembles that of a Medieval guild in
which expensive skilled artists labor extensively at one piece
at a time.
Models
and practices taken from biology change the situation dramatically.
I call the application of biological methods to the generation
of digital objects (in this case 3D but not restricted to
3D) the 'Metaforms' methodology (1996.) Rule-based software
can be used to recursively 'grow' 3D objects automatically
in large numbers. These rules are composed of two types of
symbols - transform operators, and variables which are placeholders
for later replacement with 3D component parts according to
a prescribed mapping. The rules are applied recursively on
the string of symbols describing structure of the object at
that point in time. Thus the rule-based generative system
describes the development of the object structure in four
dimensions - not only in space but also in time.
The
procedure occurs two stages. The first stage is an abstract
description of the growing structure itself, the genotype
in biological terms. At the end of some number of iterations
the string is a complete abstract description of a 3D object.
The second stage is to replace the abstract variables in the
string with 3D components which then creates a particular
intance of the genotype, i.e. a phenotype. By varying the
mapping a very diverse set of phenotypes can be created from
a single genotype. Similarly, by varying the rules or by using
stochastic rules, a large set of genotypes can also be generated.
Thus the methodology can grow huge combinatoric combinations
of component-mapped phenotypes.
The
final stage in the methodology is to use genetic algorithms
to 'breed' promising individuals, or mutate genotypes to introduce
further variations. In some cases a metric may be constructed
to judge the 'fitness' of the individuls to pass on their
genes to future generations. This enables a completely automated
system for defining growing and breeding digital content,
and in particular 3D content. In the case of aesthetic judgements
the advancement decisions are probably best made by human
discernment. However, even in this case the methodology is
very fast and efficient and hence increases and speeds digital
content production and lowers it's cost.
Mark Rudolph - Biography
Mark
Rudolph is a Java designer and interactive 3D artist and directer.
He worked at Silicon Graphics in California as a VRML director
and Java designer, and at AT&T Bell Labs doing
Multimedia software research. In September he will become
head of the Virtual Reality Center at the IT-C University
in Copenhagen. He also has a Phd in Mathematics in Transfinite
Set Theory.
Mark
Rudolph PhD.
Lucid Actual
Montreal
mrudolph@total.net
http://www.lucidactual.com
|
