The field of computer vision is witnessing a transformative shift with the advent of versatile “generalist” models trained on vast, unlabeled datasets, displaying adaptability to new tasks with minimal supervision. In the 2024 keynote, Professor Lee will present groundbreaking research on creating intelligent systems that comprehend our multimodal world with minimal human supervision, using images and text—and other modalities like video, audio and LiDAR. Attendees will gain insights into efficiently repurposing existing foundation models for specialized tasks, addressing training cost challenges and unlocking the potential of multimodal machine perception in diverse applications.