Researchers have developed Dr. CaBot, a medical AI agent that diagnoses illnesses and provides reasoning for its conclusions based on thousands of clinical case studies. Compared to human internists, Dr. CaBot exhibited significantly higher diagnostic accuracy, ranking the correct diagnosis first 60% of the time, in contrast to the 24% success rate of the physicians evaluated. The system’s effectiveness is rooted in its training on over 7,000 clinicopathological conference reports, which allowed it to adopt a consistent expert reasoning style. Additionally, in evaluations, internal medicine doctors rated Dr. CaBot’s explanations of diagnoses more favorably than those given by human peers, often unable to distinguish the AI’s outputs from those of actual doctors.
Dr. CaBot AI surpasses human doctors in diagnostic accuracy, reasoning
