Skip to main content

Google's AMIE AI Shows Off Real-Time Video Consult Skills

In the real world, a doctor's diagnosis often hinges on more than just what a patient says. It's the subtle cough, the way someone walks, or the grimace of pain that tells the real story. Now, a research AI called AMIE, developed by Google Research and DeepMind, is learning to read those cues in real-time video consultations—and it's pretty good at it.

AMIE is built on Google's Gemini and Project Astra architecture, using a multi-agent system to process what it sees and hears. In its latest upgrade, it can not only interpret visual and auditory signals but also guide virtual physical exams and reason through diagnoses on the fly. That makes it feel more like a human expert than a chatbot.

To put AMIE to the test, researchers ran a simulated consultation study. Volunteers played patients, interacting with both primary care physicians and the AI. The results? Professional evaluators gave AMIE high marks for its thoroughness in taking medical histories, diagnostic accuracy, treatment plan appropriateness, and overall communication. The patient volunteers also noted that video consultations felt more natural than text-based chats—more in line with how they'd expect to talk to a doctor.

But before you book an appointment with Dr. AI, know this: AMIE is still in the academic research stage. The team stresses that it needs rigorous validation and more research before it can be responsibly deployed in real clinical settings. Still, this progress paints an exciting picture for the future of healthcare, where AI could help bridge the gap between patients and providers, especially in remote or underserved areas.

Key Points

  • AMIE, a research AI from Google Research and DeepMind, has demonstrated expert-level real-time video consultation abilities.
  • It uses a multi-agent system built on Gemini and Project Astra to interpret visual and auditory cues.
  • In a simulated study, AMIE scored well on clinical skills like history taking, diagnosis, and treatment planning.
  • Patients found video consultations more natural than text-based interactions.
  • AMIE is not yet ready for real-world clinical use; more validation is needed.

As AI continues to evolve, the line between human and machine expertise blurs. For now, AMIE is a promising glimpse into a future where healthcare is more accessible, more intuitive, and maybe even a little more human—even when it's powered by code.