 Transformer-based chatbots have been increasingly used in healthcare due to their ability to generate accurate predictions from clinical vignettes. Two models, ChatGPT and Foresight, were compared on the task of predicting relevant diagnoses. Both models showed promising results with ChatGPT outperforming Foresight in terms of accuracy. However, there are several limitations to using these models in clinical settings, including the lack of understanding of the underlying mechanisms behind the model's predictions and the need for manual labeling of data. Despite these limitations, transformer-based chatbots show promise in providing accurate predictions in healthcare applications. This article was authored by Joshua O. Young, Joshua O. Young, Shelko Kraljevik, and others.