Explore the intersection of computer vision and robotics in this colloquium talk by Saurabh Gupta from UC Berkeley. Delve into the challenges of bridging 2D visual understanding with 3D world interaction for robots. Discover insights on representation learning for varied modalities, 3D scene understanding, and visual navigation. Learn how combining perception and action in robotic tasks reveals limitations in current computer vision techniques and opens new research directions. Gain valuable knowledge from Gupta's Ph.D. work on 3D scene understanding and visual navigation, supported by prestigious fellowships from Berkeley and Google.