ALife 2026 · Toronto
From The Artificial Experimentalist: Discovery and Control of Self-Organizing Phenomena with Autotelic Reinforcement Learning — Cvjetko, Hartl, Levin, Moulin-Frier & Oudeyer.
CARL is a goal-conditioned reinforcement learning agent that is able to guide the dynamics of Lenia (continuous cellular automaton) towards desired states through a sequence of local interventions. In this demo, CARL was trained to steer solitons (local self-organized patterns) towards a target direction. At each step, CARL can select a spot on the board and either inject or erase mass there. Use arrow keys or click anywhere on the maze to change CARL's goal. Try to navigate the soliton towards the blue circle.
Hover the ? badges next to the controls for more detail.
Performance. The Lenia simulation and the policy network both run on your device, on the CPU only, and are computationally demanding. If playback stutters, reduce the board size, lower the simulation speed, and set your machine to a high-performance power mode. Tested in desktop Chrome and Brave.
You are acting now. CARL is idle; on the board, left-click adds mass, right-click removes it and space advances one step (the Step button does the same).