Levine predicts that within five years, general-purpose robots could handle most blue-collar tasks and potentially manage homes autonomously by 2030. He explains how robotic foundation models, inspired by LLMs and vision-language models, are being developed to enable robots to learn from experience and generalize across tasks. Key challenges include improving data efficiency, model robustness, and hardware scalability. The discussion also covers the importance of simulation for training, the potential for brain-inspired computing to enhance efficiency, and how robotics could accelerate AI development. Finally, the conversation touches on global manufacturing challenges, particularly China’s role in hardware production, and the societal implications of widespread automation.