Back

GPT-4, an artificial intelligence large language model,exhibits high levels of accuracy on dermatology specialty certificate exam questions.

Shetty, M.; Ettlinger, M.; Lynch, M.

2023-07-14 dermatology
10.1101/2023.07.13.23292418 medRxiv
Show abstract

Artificial Intelligence (AI) has shown considerable potential within medical fields including dermatology. In recent years a new form of AI, large language models, has shown impressive performance in complex textual reasoning across a wide range of domains including standardised medical licensing exam questions. Here, we compare the performance of different models within the GPT family (GPT-3, GPT-3.5, and GPT-4) on 89 publicly available sample questions from the Dermatology specialty certificate examination. We find that despite no specific training on dermatological text, GPT-4, the most advanced large language model, exhibits remarkable accuracy - answering in excess of 85% of questions correctly, at a level that would likely be sufficient to pass the SCE exam.

Matching journals

The top 12 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.