DIFFERENT
BITS
ARTIFICIAL LIFE – PART 3
GENETIC ALGORITHMS: INTERACTIVE EVOLUTION
In the previous two parts of this series, we examined the theory behind Genetic
Algorithms and implemented a toy “Hello World” example on the PIC microcontroller.
As a quick reminder, Genetic
Algorithms are computer
processes that draw inspiration
from the natural process of evolution.
They are used both to evolve answers
to difficult computational problems
and to simulate the historical
progression of species. Genetic
Algorithms evolve populations of
candidate solutions, which are then
evaluated on the basis of fitness. Their
fitness rating is derived from their
proximity to the goal state, and they
breed based on this rating and a
degree of randomness. Utilizing
methods of cross-over and mutation,
Genetic Algorithms are able to find
answers to problems that are too
computationally expensive to tackle
with brute force or more traditional
heuristic searches. The mathematical
possibilities are exciting, but Genetic
Algorithms can also be beautiful and
it is this artistic and aesthetic potential
we will be exploring in this article.
First off, you might be thinking:
“How can an algorithm be beautiful?”
So, a few examples and a little bit of
history is in order.
Psychologist, electrical engineer,
and computer scientist John Holland
first formalized Genetic Algorithms
in 1975. The link that he so aptly
captured between biology and
computation soon inspired many
others, including programmers, visual
artists, and musicians. The idea of
evolving art, however, or a piece of
music introduces special problems that
the original Genetic Algorithm does
not attend to: How can a computer
evolve a conclusion when there is no
“right answer;” when the criterion for
success is entirely subjective?
Appropriately, it was an artist named
William Latham and a programmer
named Stephen Todd in
the burgeoning field of
computer graphics who
solved this problem,
creating software
for evolving three-dimensional sculptures
in the computer. The
work was inspired by
Richard Dawkins’ book
The Blind Watchmaker
in which he discusses
experiments with a
computer program he
wrote called “Biomorph
Land” which allowed
him to guide the
evolution of stick figures
over several generations. Although
Dawkins’ goal was to show how
natural selection and the accumulation
of small changes over time could
account for the diversity of life on
earth, others who read his book —
like Latham and Karl Sims — became
inspired by the visual images the work
evoked. Sims was the first to bring
these ideas out into the public in a
1993 installation at Ars Electronica
called “Genetic Images.” The piece
“Safari” of biomorphs
created by Richard Dawkins’
software described in
The Blind Watchmaker.
SERVO 09.2008 67
