Advancements in learning language models (LLMs) are raising questions about their potential benefits to society, particularly in the medical field. A recent study conducted by the University of Cambridge’s School of Clinical Medicine revealed that OpenAI’s GPT-4 demonstrated impressive performance in an ophthalmology assessment, nearly matching that of expert practitioners in the field.
Published in PLOS Digital Health, the study compared GPT-4 with its predecessor GPT-3.5, Google’s PaLM 2, and Meta’s LLaMA by administering 87 multiple choice questions related to ophthalmology. A diverse group of participants, including expert ophthalmologists, trainee ophthalmologists, and junior doctors, took the exam. The questions covered a range of topics from light sensitivity to lesions, sourced from a specialized textbook. Since the content was not publicly available, researchers believe that the LLMs were not trained on these specific questions. ChatGPT, powered by GPT-4 or GPT-3.5, was given multiple opportunities to provide accurate responses.
According to the Financial Times, the results of this study highlight the potential of advanced LLMs like GPT-4 to assist medical professionals in diagnostic assessments and decision-making processes. The implications of these findings could revolutionize the way healthcare professionals leverage AI technology in their practice.
2024-04-18 08:15:17
Source from www.engadget.com