This chapter outlines future directions in robotics, emphasizing the importance of building strong robot foundation models by framing robotics as a visionlanguage problem. It highlights key challenges such as motion generalization, scaling visual-language-action (VLA) models, and ensuring safety and alignment for real-world deployment. The chapter envisions a future where semi-autonomous systems learn through scaled interaction and human-in-the-loop feedback to achieve general-purpose intelligence in physical environments.