Researchers have demonstrated the first expert-level AI system for real-time clinical video consultations using AMIE (Video), a Gemini-based multi-agent system that integrates low-latency dialogue with real-time audio-visual perception. In a randomized Objective Structured Clinical Examination study involving 30 primary care physicians and 100 scenarios, clinical evaluators rated the system on par with or better than physicians in history-taking, diagnosis, management, and physical observation.

  • AMIE (Video) outperformed its text-only counterpart in patient preference for communicative effectiveness and feeling understood.
  • Primary care physicians were still preferred by patients for rapport and partnership building.
  • The system showed limitations in fine anatomical precision, subtle affective nuances, and high-frequency movements.

These results mark a milestone toward AI systems capable of augmenting care across the sensory complexity of clinical practice, though further research is needed before real-world translation.