As ChatGPT makes its way into healthcare, researchers at Evanston, Ill.-based Northwestern University recently demonstrated three uses for large language models in medicine.
Alex Carvalho, MD, an infectious disease fellow at Northwestern's Feinberg School of Medicine, asked ChatGPT how it would treat a real clinical case: a 39-year-old gunshot victim who developed an infection, according to the June 13 news release. It made what he believed to be the wrong decision initially but the right one once he gave the artificial intelligence chatbot additional details. David Liebovitz, co-director of the Institute for Augmented Intelligence in Medicine and the Center for Medical Education in Digital Healthcare and Data Science, created a template that encouraged ChatGPT to link to a trusted medical database in its responses.
Two computer science students who are part of the university's Cognition, Creativity and Communication machine learning lab created an analytics engine using emergency department admissions data that it feeds to the AI. "The language model will provide the structure and the fluency, but the facts themselves are coming from the data," said Kristian Hammond, PhD, a computer science professor and director of the Center for Advancing Safety of Machine Intelligence, in the release.
The third demonstration used large language models to create AI assistants to educate low-literacy patients about their health issues. The researchers found that how the avatars looked played an important role in whether they were effective.