In this episode of NEJM AI Grand Rounds, hosts Raj Manrai and Andy Beam interview Professor Emily Alsentzer from Stanford University, an expert in natural language processing and large language models. Emily discusses her background, from her early exposure to medicine through her family to her interest in computer science sparked by the book "The Most Human Human." She shares insights into her work with clinical BERT embeddings, questioning the necessity of specialized clinical models in light of advancements like GPT-4, and evaluating the potential for racial and gender biases in GPT-4 within healthcare settings. The conversation explores the balance between leveraging large language models and maintaining rigorous evaluation standards, as well as the importance of interdisciplinary training in AI and medicine.
