Avoiding Side Effects in Complex Environments
This addresses the challenge of reward function specification for AI agents in complex environments, though it is incremental as it builds on prior work in toy settings.
The paper tackles the problem of avoiding negative side effects in reinforcement learning agents by scaling the Attainable Utility Preservation (AUP) method to large, randomly generated environments based on Conway's Game of Life, resulting in modest overhead while enabling task completion and avoidance of many side effects.
Reward function specification can be difficult. Rewarding the agent for making a widget may be easy, but penalizing the multitude of possible negative side effects is hard. In toy environments, Attainable Utility Preservation (AUP) avoided side effects by penalizing shifts in the ability to achieve randomly generated goals. We scale this approach to large, randomly generated environments based on Conway's Game of Life. By preserving optimal value for a single randomly generated reward function, AUP incurs modest overhead while leading the agent to complete the specified task and avoid many side effects. Videos and code are available at https://avoiding-side-effects.github.io/.