Artificial Intelligence vs. ID Residents and Specialists: Who Tested Best?

While GPT-4's theoretical knowledge was good, it performed less well in clinical scenarios requiring nuanced judgment.

ChatGPT, the software that uses the GPT family of large language models (LLMs), has enabled widespread use of artificial intelligence (AI) in many domains, including medicine. GPT-4 has demonstrated expert-like performance in general medicine cases, but its ability to solve more-specialized clinical problems is less certain.

To find out how AI performs for clinical problem solving in infectious diseases (ID), researchers in Italy asked four experienced ID specialists, four senior ID residents, and ChatGPT (standard version and trained version supplemented with international guidelines, clinical trials, systematic reviews, and meta-analyses) to responded to six true/false questions and six open-ended questions, and to solve six clinical cases…

Reviewing Authors

Thomas Glück, MD

Disclosures

Editorial BoardsConsilium Infectiorum by InfectoPharm, Infection

Thomas Glück, MD

Disclosures

Editorial BoardsConsilium Infectiorum by InfectoPharm, Infection

Raja-Elie E. Abdulnour, MD

Disclosures

Nothing to disclose

Raja-Elie E. Abdulnour, MD

Disclosures

Nothing to disclose

Artificial Intelligence vs. ID Residents and Specialists: Who Tested Best?

Topics

Reviewing Authors