Building General Zero-Shot Robot Policies - How Large Models Changed Robotics

Watch a seminar presentation from MIT exploring the impact of large language models on robotics and the path toward general, zero-shot robot policies. Discover how traditional Task and Motion Planning methods can be enhanced with machine learning to improve real-world generalization. Learn about groundbreaking developments including PaLM-E for embodied decision making, RT-2 for connecting language models to robot actions, and Aloha Unleashed for advancing robot dexterity. Gain insights into the integration of small-scale machine learning with traditional robotics approaches, and understand the transformative effects of large models on the field. Explore perspectives on fundamental research challenges that remain and potential future directions for robotics research.