This project consisted of two phases, and was an attempt to build an autonomous art system that could learn about the physical world in which it was situated.
The installation occupied two gallery spaces, one with four robots in it and the other with people in it and computer terminals for them to remotely communicate with the robots. Sitting at the terminals, people would be asked to provide a basic profile of themselves to the system: interests, confidence, happiness, etc. The system would then assess your eligibility as a “parent” for one of the robots in the other room. If it liked the look of you, then it would extract another, virtual co-parent from its database of previous successful parents and between you, a genetic profile would be created that was then downloaded into one of the robots in the other room.
The result would be a robot that had a personality constructed from its two parents, and the personality would determine the behavior, characteristics and goals that the robot would have during its life span. You could then make suggestions to the robot about what it did: which direction it should move in, how fast it should move, how confident it would be in approaching other robots, how sociable it would be, how adventurous it would be, how shy it would be. So the robots would investigate the layout of the other room and the other robots in it. The robots were also designed to be able to learn from each other: when they came into proximity of one another, their personality traits would be beamed across using short-range infrared. Everything that the robots did would be communicated back to a central database, which was running a series of scripts coded using genetic algorithms. The scripts would assess whether the robots were learning new things from their parents. If the information being harvested by the robots from their parents fell below a particular threshold, the system would immediately disconnect the parent from its robot. Following that it would wait until it found another parent from those typing in their profiles, who had something to offer that represented new information for the system. The nature of this threshold, and the things that the system prioritized – in other words the things that it wanted to learn – was dynamically evolving due to the genetic algorithms.