AI are better than we are at guessing a negative diagnosis

Embargoed until: Publicly released:

US researchers pumped medical diagnostic questions into ChatGPT and compared the AI’s answers to those of over 550 practitioners. The researchers found the AI was more accurate than humans in determining a negative test result both before and after the test, but humans still outranked the bot at working out a positive result.

Journal/conference: JAMA Network Open

Link to research (DOI): 10.1001/jamanetworkopen.2023.47075

Organisation/s: Beth Israel Deaconess Medical Center, USA

Funder: No specific funding listed


Note: Not all attachments are visible to the general public

  •  Research


    Web page


Leave a Comment