Google AI better than human doctors at diagnosing rashes from ...
Google's medical chatbot, powered by artificial intelligence (AI), has proven to be more accurate than human doctors in diagnosing rashes. This advanced chatbot can analyze patient information and photos to identify the underlying cause of a rash, among other medical conditions.
The upgraded version of Google's medical chatbot can now utilize smartphone photos to diagnose rashes, as well as assess various types of medical imagery. In a recent study, the AI system outperformed physicians not only in diagnostic accuracy but also in bedside manner. It also excelled at interpreting medical images like electrocardiograms and laboratory results.
Experimental AI System: Articulate Medical Intelligence Explorer (AMIE)
The latest iteration of the AI system, known as Articulate Medical Intelligence Explorer (AMIE), is still in the experimental phase. Described in a paper posted to the arXiv preprint server, this system has not yet undergone peer review. The development of systems like AMIE, which integrate images with clinical information, represents a significant step towards creating an AI assistant that mimics a clinician's thinking process.
Evaluating Diagnostic Capabilities
To assess AMIE's diagnostic capabilities, a study involved 25 actors simulating patients in virtual consultations with both the chatbot and a human primary-care physician. These actors presented a variety of medical scenarios, symptoms, medical histories, and relevant images for evaluation.
After each consultation, AMIE and the human doctor proposed diagnoses and treatment plans. Specialists in dermatology, cardiology, and internal medicine evaluated the performances of the chatbot and human doctors based on transcripts of the visits and post-visit reports. Overall, AMIE demonstrated higher accuracy in diagnosing conditions compared to human doctors, with its performance being less affected by issues like low-quality images.
Advanced AI Technology
The updated chatbot is based on Google's large language model (LLM) Gemini 2.0 Flash, which has been tailored for medical purposes. By enhancing the LLM's ability to engage in diagnostic conversations and clinical reasoning, researchers have improved the system's efficiency in analyzing medical data.
Researchers confirmed the superiority of the upgraded system by simulating patient-physician dialogues, where the AI model assumed the roles of a patient, a physician, and an evaluator of their interactions. This approach helped instill desirable behaviors in the AI during diagnostic conversations.
Citations: Source
