Yale NLP/LLM Interest Group

Foundation models (FMs) have emerged as transformative tools in AI-driven healthcare, promising unprecedented capabilities for biomedical data analysis. In this talk, Xiaoxiao Li, PhD will first provide an overview of our recent advances in incorporating transformer-based FMs into medical image analysis. Despite their successes, medical FMs face critical challenges, including biases, limited interpretability, insufficient adaptability to clinical contexts, etc. As the focus of this talk, she will share our exploration of these challenges, emphasizing strategies that significantly enhance fairness, robustness, and clinical utility. Additionally, Li will share her in-depth analysis on critically examining the ongoing debate over whether general-purpose foundation models are justified, considering their substantial cost, or if specialized medical foundation models alone suffice. Lastly, Li will outline her strategies to unlock the full potential of trustworthy and impactful AI in healthcare.

-------

Xiaoxiao Li, PhD is currently an Assistant Professor in the Department of Electrical and Computer Engineering at the University of British Columbia, a faculty member at Vector Institute. Li is recognized as a Canada Research Chair (Tier II) in responsible AI and a Cifar AI Chair. Li’s research interests primarily lie at the intersection of AI and healthcare, theory and techniques for artificial general intelligence (AGI), and AI trustworthiness. Li aims to develop the next-generation responsible AI algorithms and systems.