Giving live to autonomous agents

Autonomous agents created using machine learning are a growing presence in our world, so we should use them for art! Participants of this workshop will livecode virtual worlds and populate them with intelligent agents.

The workshop will start with a brief overview of tools and terminology. Training (outside the scope of this workshop) is accomplished using Unity’s ml-agents and simplified with PersonalityQuarks. The primary livecode interface is Arcadia with a few helpful wrappers from baisless fabric.

The first half hour - hour will be the overview and setting up tools. Participants should download Unity and a github repository with all the important parts set up (link to be provided later), but there’s always going to be complications.

The remainder of the time will be spent on exploring the environment. Participants will start by creating an agent prefab and then spawning it. Then they’ll experiment with multiple agents, then modifying the environment for agents to interact together, and finally (if there’s time) customizing their environment with colors and lights. By the end of the workshop, participants should have an understanding of how to use autonomous agents in a live performance, and have the tools to train their own agents and create their own environments.
This will be a moderately advanced workshop, and participants will preferably have some familiarity with lisp and 3D environments.