Researchers proposed a reinforcement learning method to bias perception in AI models. It uses a Complementary Learning System with separate short-term and long-term memory. The approach improves few-shot learning and reduces forgetting in continual learning. Tests on image classification showed better resistance to forgetting than a ResNet baseline. Few-shot generalization also outperformed other continual learners. Overall accuracy in continual learning was slightly lower than existing models. The limitation stems from the long-term memory's representations and plasticity. Future work will replay short-term memories to the long-term memory. The paper is 21 pages with 8 figures.