Open Ended Ensemble (Competitive Coevolution)

Open Ended Ensemble is a series of installations in which a collection of electromechanical sound-producing automatons, or agents, integrate basic attributes of a biological living system. In the current version (Competitive Coevolution), an ensemble of self-modifying, high-voltage sound amplifiers slowly learn to maneuver a robotic probe to the weakest region of electro-magnetic radiation along a fluorescent light fixture. Strong radiation from the lights causes the amplifiers to produce sound, layering sympathetic tones over the familiar drone of fluorescent lighting.

The agent’s behaviour is adaptive, subject to an evolutionary process in which a random population of computer programs slowly evolve, eventually achieving enough control of the robotic probe to coax its movement away from the source of radiation and into silence. Meanwhile, the light fixture would prefer to maintain the drone, and slowly evolves a strategy of its own, learning to move the lights and trap the probe in regions of strong radiation. An arms race ensues as the two competing forces interact and coevolve, akin to predator/prey or host/parasite relationships in biological systems.

Agents in the Open Ended Ensemble are interacting with simple, conflicting goals in a never-ending game. Their sensory experience of the environment is inherently noisy, obtained entirely from the coupling of a bare-bones magnetic probe to a fluorescent light fixture. Their control of the probe is very imprecise and clumsy. As a result, the agents are navigating their world with partial information and limited motor control. A volatile model ecosystem emerges in which each device achieves intermittent success and failure as their interaction unfolds over days or weeks.

This installation was developed under the scope of Peripheral Visions, curated by Eryn Foster for Nocturne 2014 in Halifax, NS and Slow Now Constant, curated by Ryan Suter for Heat Wave in Sackville, NB.

Thank you: Arts Nova Scotia, Eryn Foster, Nocturne Halifax Team, Ryan Suter, Struts Gallery & Faucet Media Arts Centre.

On Musical Automatons and Biologically-Inspired Design
Musical automatons are among the earliest self-operating machines. In the Middle Ages, artists such as al-Jazari crafted complex mechanical systems, essentially programmable robots, that could perform music. Later, Jaquet-Droz created The Musician, a mechanical android resembling a young girl that plays a real piano, and the Singing Bird Box, a small box that contained a miniature singing bird. These automatons aimed to mimic the behaviour of living creatures, either humans or animals, and were often successful enough to be temporarily mistaken for the real thing by audience members. However, their actual resemblance to living systems is superficial. Most historical automatons are elaborate parlour tricks, illusions operated by intricate clockwork. Their behaviour is predetermined and predictable. In contrast, the Open Ended Ensemble approaches the development of musical automatons with a focus on introducing attributes of biological living systems, disregarding any attempt to superficially mimic real-life scenarios such as birds singing or humanoid musicians. The attributes of living systems I consider key to this process are: 1) Modularity 2) Adaptability, and 3) Self-replication. Thus, instead of setting out with the objective to create a machine that looks like a bird and accurately reproduces bird-like songs, the objective here is to develop electromechanical sound machines that are complex systems made from multiple building blocks (modularity), capable of adapting their behaviour, physical structure, and sonic properties based on interaction with their environment (adaptability), and able to reproduce by making physical copies of themselves (self-replication). The title Open Ended Ensemble refers to the idea that sounds produced by these machines, as well as their behaviour and physical characteristics (eg. circuit board design), are emergent properties that evolve over long periods of time (weeks, months). While individual elements of the work may have a robotic visual aesthetic, embracing biologically-inspired design principals might result in the overall work behaving like a natural system.

2014

Open Ended Ensemble (Reinforcement Learning)