Explore unsupervised biomedical image segmentation using hyperbolic representations in this Stanford University lecture. Delve into Jeffrey Gu's research on leveraging inherent hierarchical structures in biomedical images to train segmentation models without labeled datasets. Learn about the novel self-supervised hierarchical loss and the advantages of hyperbolic representations in capturing tree-like structures. Gain insights into the application of these techniques in biomedical imaging and their potential impact on the field. Discover the speaker's background, research interests, and the importance of this work in advancing unsupervised learning for medical image analysis. Engage with topics such as brain tumor imaging, machine learning methods, self-supervised learning, and evaluation techniques. Participate in the discussion on future work and potential applications of this innovative approach to biomedical image segmentation.