Back to:
last page

Prix 1987 - 2007

ORF Oberösterreich

Jellylife / Jellycycle / Jelly Locomotion
Michael Joaquin Grey

"Jellylife" by Michael Joaquin Grey is a still image clipped from a 3-D animation showing the development of genetic algorithms.

I have included three highly related works, "Jellylife", "Jellycycle" and "Jelly Locomotion". They are 3-D animations of the development of neural networks and genetic algorithms. They were processed by a Stardent supercomputer. We designed the neural nets and genetic algorithms ourselves, the visualisation was on scientific visualisation software designed by Stardent. The cels were printed on a Tectronics printer. The behaviour of the information is visualised here as animation cels, graphite tablets, and 3-D stereo lithographic models. The models are solidified by drawing the 3-D files with a UV laser in a liquid UV sensitive resin, a process developed by Quadrax Corp.

As an example, "Jellycycle" has over a quarter of a million stages in its developmental cycle, enough to present it as a full length film, "à la cinéma verité", however we choose to present it as 56 stages, each a still image (so as to be able to see the complete development of the cycle all at once). All these pieces are both computer images and computer animations.

The images represent 3-D forms, self-organising systems, representing crucial morphological changes in the growth of the networks. The network is mapped onto a simple form to begin with, and its growth is dependent on the variation in the environment (random number generator). The supercomputer supports our random number generator which feeds the neural networks (the number generator is partly based on the computer clock, it could, however, be any data generator; it is random only because it attempts to create a control model for observing variation in form and behaviour). The unique development of the networks is created by the variation in the random number generator, and the design of the networks and algorithms. The neural networks and genetic algorithms are uniquely, able to "cope" and "learn" and "live" in this robust environment. At best, we were attempting to be early observers and describe the behaviour of the information on a scale of magnitude only recently beginning to be explored. As early observers and commentators on new behaviour and form (in contrast to the inevitable explainers and exploiters), we took the historical models of Leeuwenhoek, Audubon or Kepler and considered consciously our own biases andnarrative in discovering "Artificial Life".

The three pieces chosen exhibit three specific behaviours. "Jellycycle" is a 3D neural network which has a full cycle of close to 254,000 repetitions. It develops from a single sphere to a highly altered self-organising form. The ontogeny of information mimics many stages of early development of marine invertebrates. The actual 3D model, "Jellylife", has a highly complex internal anatomy; it is divided into 4 stages and holds over a gigabit of memory for the single 3D image and sculpture. 3 x 16 Syntax (Palindrome) and Carbon Codex 16 x 16 are both a genetic algorithmand neural network. 3 x 16 Syntax (Palindrome) is an excerpt of the Carbon Codex, 3 of the 16 lines created in the time period (T:June 6,1992,12:06 to 12:32 PM). Each repetition represents one character, and there is a 16 character long syntax which is repeated 16 times. Every cycle, the character evolves and becomes more "gothic" in its structure, with increased variegation and branching off of the main stalk of the character. The 16 characters seem to be consistently present asa palindrome 8, then 8 backwards, stereoscopic "isomers" or 3-D "mirror" images. A palindrome is a formal syntactical strategy employed by actual genetic material to preserve its tertiary structure when being decoded to reduce translation error.

"Jelly Locomotion" ("Cancerhead") is a fully developed neural network which wanders about, due to a slight bias towards a certain not so random number generator. The 3D self organising system, which seems to very much resemble ajellyfish or medusa, is visualised from three vantage points: above, straight portrait, and below; the presentation of its motion was visualised to mimic the classic Muybridge presentation of behaviour and movement.

Technical Background
HW: Stardent Supercomputer
SW: Artist's Proprietary (Genetic Algorithms)