Keeping this in mind, a tool that is correct 76% of the time could also prove valuable. Someone living in a remote area or facing financial barriers to healthcare may benefit from information that helps them understand symptoms, identify warning signs, or determine whether urgent medical attention is necessary. However, the researchers also found that AI models struggled most with questions related to dermatology, mental health and internal medicine. Dermatological issues often require visual assessment, something current AI systems still perform less effectively than text-based tasks.
