Join the Natural Language Processing Working Group for an energetic discussion of the article Performance of ChatGPT on USMLE: Potential for AI-assisted medical education using large language models by Tiffany Kung et al. The authors will review the explosive modern history of biomedical domain-specific LLMs for MedQA tasks, present new data showing unprecedented zero-shot performance of pretrained foundation (general purpose) LLM GPT-3.5 on the USMLE, and tour several case studies highlighting the pertinent success and failure modes. Finally, the conversion will explore issues of trust, accuracy, and explainability when evaluating emerging AI technologies as copilots in healthcare.
Presenters
Victor Tseng’s research program at AnsibleHealth focuses on improving polychronic care delivery for complex conditions using digital health and clinical AI. Active areas of investigation include human-AI alignment for CDS, real time risk management, and dimensionality reduction in chaotic environments.
Dr. Tseng is a practicing internist, pulmonologist, and intensive care physician. His clinical areas of interest include chronic respiratory failure, transitions of care, and systemic impacts of chronic lung disease. He is also a faculty member and medical educator at UWorld, LLC, where he supervises broad curriculum development for undergraduate, postgraduate, and CME learners.
Dr. Tseng earned his medical degree from the University of Wisconsin School of Medicine & Public Health. His internship and residency were completed at Emory University, followed by fellowship training in Pulmonary Sciences & Critical Care Medicine at the University of Colorado/National Jewish Health combined program.
Tiffany Kung is a resident physician in anesthesia at Massachusetts General Hospital. She earned her medical degree from Stanford School of Medicine. Dr. Kung worked at Google in the healthcare AI department for two years. Her clinical areas of interest include global health, medical education, and affordable medical technology.