Recent research from the University of Milan explores whether artificial intelligence (AI) can replace doctors in diagnosing neurological conditions. The study evaluated the performance of popular AI models, ChatGPT and Gemini, during initial neurological examinations involving 28 patients. While neurologists achieved a diagnostic accuracy of 75%, ChatGPT and Gemini managed only 55% and 46%, respectively. Both AI models exhibited a tendency to over-prescribe diagnostic exams in 17–25% of cases. Despite showing potential as support tools, the findings suggest that AI cannot currently replace human evaluation in complex fields like neurology. The researchers advocate for responsible integration of AI in clinical settings, emphasizing the importance of human oversight. They highlight the need for tailored training in AI for medical students. Ultimately, the study indicates that while AI tools like ChatGPT and Gemini can aid clinical practice, rigorous human supervision and further research are essential for safe adoption.
Source link
