AI in healthcare is a hot topic, and Stanford Health’s Data Science team is leading the conversation. Recently, Nigam Shah, Chief Data Scientist and Professor of Medicine and Biomedical Data Science, and Jason Fries, Research Scientist at the School of Medicine, moderated a dialogue on evaluating large language models (LLMs) like ChatGPT. They discussed the pros and cons of foundational models, and how to best assess their use in healthcare. The advantages of such models include better accuracy, simplified deployment, emergent clinical abilities, and novel human-AI interfaces. But there are drawbacks too, such as a lack of shared evaluation frameworks and datasets. Ultimately, both Shah and Fries believe that AI can augment existing roles and provide valuable insights, while humans must remain in the loop. AI and healthcare are two dynamic fields, and it’s great to see the strides being made to ensure responsible AI development and use.

Read the full publication ››