Do We Still Need Clinical Large Language Models?

Explore the utility of specialized clinical language models in a 27-minute conference talk from the Toronto Machine Learning Series. Delve into an extensive empirical analysis comparing 12 language models of varying sizes on three clinical tasks. Discover how smaller domain-specific models outperform larger general-purpose language models in highly specialized, safety-critical medical contexts. Learn about the benefits of pretraining on clinical tokens for creating more parameter-efficient models. Gain insights from Alistar Johnson, a Scientist at SickKids, on the potential advantages of focused clinical models over general large language models in healthcare applications.