Microsoft Bing outperforms five other generative artificial intelligence chatbots in the Antwerp University multiple choice medical license exam

Abstract

Abstract Recently developed chatbots based on large language models (further called bots) have promising features which could facilitate medical education. Several bots are freely available, but their proficiency has been insufficiently evaluated. In this study the authors have tested the current performance on the multiple-choice medical licensing exam of University of Antwerp (Belgium) of six widely used bots: ChatGPT (OpenAI), Bard (Google), New Bing (Microsoft), Claude instant (Anthropic), Claude+ (Anthropic) and GPT-4 (OpenAI). The primary outcome was the performance on the exam expressed as a proportion of correct answers. Secondary analyses were done for a variety of features in the exam questions: easy versus difficult questions, grammatically positive versus negative questions, and clinical vignettes versus theoretical questions. Reasoning errors and untruthful statements (hallucinations) in the bots’ answers were examined. All bots passed the exam; Bing and GPT-4 (both 76% correct answers) outperformed the other bots (62-67%, p= 0.03) and students (61%). Bots performed worse on difficult questions (62%, p= 0.06), but outperformed students (32%) on those questions even more (p<0.01). Hallucinations were found in 7% of Bing’s and GPT4’s answers, significantly lower than Bard (22%, p<0.01) and Claude Instant (19%, p=0.02). Although the creators of all bots try to some extent to avoid their bots being used as a medical doctor, none of the tested bots succeeded as none refused to answer all clinical case questions. Bing was able to detect weak or ambiguous exam questions. Bots could be used as a time efficient tool to improve the quality of a multiple-choice exam. Author Summary Artificial chatbots such as ChatGPT have recently gained a lot of attention. They can pass exams for medical doctors, sometimes they even perform better than regular students. In this study, we have tested ChatGPT and five other (newer) chatbots in the multiple-choice exam that students in Antwerp (Belgium) must pass to obtain the degree of medical doctor. All bots passed the exam with results similar or better than the students. Microsoft Bing scored the best of all tested bots but still produces hallucinations (untruthful statements or reasoning errors) in seven percent of the answers. Bots performed worse on difficult questions but they outperformed students on those questions even more. Maybe they are most useful when humans don’t know the answer themselves? The creators of the bots try to some extent to avoid their bots being used as a medical doctor, none of the tested bots succeeded as none refused to answer all clinical case questions. Microsoft Bing also turns out to be useful to find weak questions and as such improve the exam.

Ähnliche Arbeiten