Healthcare systems globally are working toward improved outcomes, reduced costs, and a better experience for patients and clinicians alike. However, progress is limited by a worldwide shortage of clinical specialists. The World Health Organization forecasts a deficit of over 10 million healthcare workers by 2030. Although AI is widely viewed as a solution to this gap, it has yet to fully address the requirements of clinicians and patients. Today, we’re announcing our AI co-clinician research initiative to investigate how AI can enhance physicians‘ expertise and improve the quality of care delivered to patients. At Google DeepMind, our work in medical AI has progressed from achieving expert-level performance on medical knowledge exams with MedPaLM, to matching doctors‘ capabilities in text-based simulated consultations with AMIE — including in real-world clinical feasibility studies. We also have a long track record of researching how clinicians and AI can collaborate. We hypothesize that the future of healthcare will involve „triadic care,“ in which AI agents support patients throughout their care journey under the supervision and clinical authority of their physician. Medicine has always been a team effort, and AI agents can expand that team by extending clinicians’ capabilities while preserving their judgment and oversight. This principle forms the core of our AI co-clinician research program: developing AI that acts as a true collaborative partner within the care team, engaging with patients only under the supervision of expert clinicians. We developed and assessed an AI co-clinician in both clinician-facing and patient-facing environments. The key to improving the quality, cost, availability, and overall experience of healthcare lies in addressing both perspectives. Continued research is needed to make medical AI systems more trustworthy and genuinely useful to clinicians as they care for patients. Ultimately, the goal is to augment clinicians with reliable AI co-clinicians. For a doctor, a tool is only valuable if it can be trusted and is based on solid facts.
Google DeepMind News