Submitter: George Simms, I-Dat, hello@georgesimms.net
Abstract:
This project is an exploration into AI’s abilities to form visual representations of its own complex relationships. The work manifests as a multi-agent reinforcement learning environment, most notably affording the agents within, the ability to represent themselves to themselves and each other. In doing this the project aims to explore the material semiotics of these relationship builds (AI) allowing them to compose and re-present their relations with each other and their environment through a visual level. The agents’ environment is simple, each spherical agent must not collide with one another or fall off the plane, allowing for the core element of their sensed inputs to be reconfigured. Instead of using accurate metrics of their environment and surroundings, they are allowed to dictate their representation of it to themselves through an abstract shader, entangling each agent through their representational parameters into one homogenous imag(in)ing. The current stage of development is starting to introduce different roles within the agents to create a constructive conflict within the environment, trying to induce more complex and diverse performances of the agents’ relations.
Areas of STS Scholarship: Information, Computing and Media Technology, Feminist STS, Method and Practice
