Moloch is a simulation that follows the journey of a Reinforcement Learning agent as it learns to navigate through it’s virtual environment. Reinforcement Learning belongs to a Machine Learning framework where autonomous entities, typically called agents, learn how to do specific tasks through a process of trial and error, and guided by rewards and punishments in pursuit of the optimal behaviour. These agents possess a set of actions and observations about their virtual environment, training for an episode at a time, in simulations that are not limited to our time scale. In this work, the training process is transformed into a computational ritual of sacrifice, learning from the mistakes of its past. The simulation is slowed down to a smaller time-scale, observing the agent as it learns to fall, stand, and walk through a endless loop of failure.

Created as part of “The OBJECT Itself” exhibition, UCCs & Sample-Studios Digital Art in Ireland Symposium 2024. Supported by the Arts Council of Ireland, University College Cork, Sample-Studios, & Artlinks.

Built in Unity using the ML-Agents RL Framework. Sound design – Inwoo Jung.