At Upper Bound 2025, Canada CIFAR AI Chair Mo Chen asked a key question in modern robotics: why are robots still largely absent from complex, human-centric spaces like hospitals and restaurants despite advanced hardware? 

To address this, he proposes a "hybrid AI" approach that combines two approaches: the data efficiency and interpretability of classical AI (planning, control theory) with the powerful intuition and scalability of modern AI (deep learning, RL). Through case studies on human motion prediction and a "follow-ahead" robot, he demonstrates how this hybrid system allows robots to understand human intent and navigate dynamic environments.