Graceful task adaptation with a bi-hemispheric RL agent
This work addresses the challenge of task adaptation in reinforcement learning for AI agents, though it appears incremental as it builds on existing brain-inspired models.
The paper tackled the problem of poor initial performance on novel tasks in reinforcement learning by developing a bi-hemispheric agent inspired by the human brain's Novelty-Routine Hypothesis, which resulted in minimal impact on learning novel tasks.
In humans, responsibility for performing a task gradually shifts from the right hemisphere to the left. The Novelty-Routine Hypothesis (NRH) states that the right and left hemispheres are used to perform novel and routine tasks respectively, enabling us to learn a diverse range of novel tasks while performing the task capably. Drawing on the NRH, we develop a reinforcement learning agent with specialised hemispheres that can exploit generalist knowledge from the right-hemisphere to avoid poor initial performance on novel tasks. In addition, we find that this design has minimal impact on its ability to learn novel tasks. We conclude by identifying improvements to our agent and exploring potential expansion to the continual learning setting.