AI Co-clinician Research Initiative

3 posts · Apr 2026 – May 2026

What this trail covers

The AI Co-clinician research initiative continues to explore how AI can augment physician expertise and improve patient care. This involves developing AI agents that can function as collaborative members of the care team, interacting with patients under clinical supervision. Research focuses on clinician-facing support, including evidence synthesis and medication knowledge, as well as patient-facing multimodal capabilities using live audio and video for telemedical settings. Safeguards include a

All DeepMind trails

May 2026Co-Scientist: A multi-agent AI partner to accelerate researchThis post introduces Co-Scientist, a multi-agent AI system built with Gemini, designed to accelerate scientific hypothesis generation. It details the system's architecture, which includes specialized agents for idea generation, debate, and evolution, orchestrated by a supervisor agent. The 'tournament of ideas' approach, inspired by AlphaGo and AlphaStar, is used for ranking hypotheses, with a strong emphasis on verification against scientific literature and data. The system integrates web search and specialized databases, with testing of advanced models like AlphaFold. Co-Scientist is being made available to individual researchers via Gemini for Science and to enterprise partners.
May 2026Co-ScientistThis post details the application of Co-Scientist, an AI agent, in accelerating drug discovery for liver fibrosis. Gary Peltz's team used Co-Scientist to identify potential drug candidates from scientific literature. The AI proposed three candidates, two of which showed significant efficacy in lab tests, including the cancer drug vorinostat which blocked 91% of a damage response. This demonstrates Co-Scientist's ability to identify drugs that reshape gene activity, suggesting a new generation of anti-fibrotic medicines.
May 2026Co-ScientistThis post details the application of Co-Scientist, a Gemini-powered AI agent, in accelerating scientific discovery. It highlights how Co-Scientist can process grant proposals and unpublished data to generate and rank hypotheses, identify novel research targets (e.g., specific proteins and amino acids), and significantly reduce the experimental timeline from years to months. The agent's ability to refine hypotheses through iterative feedback and handle confidential data is demonstrated.
May 2026Co-ScientistThis post details the application of Co-Scientist, a Gemini-powered AI agent, to assist researchers at Calico Life Sciences in the biology of aging. Co-Scientist was used to generate a novel hypothesis about the regulation of the integrated stress response (ISR) by metabolism, which is relevant to aging and disease. The agent also assisted in refining experimental designs to test this hypothesis and incorporate emerging results. The experiments yielded new findings with implications for ISR in health and disease.
Apr 2026AI co-clinician: researching the path toward AI-augmented careThis post details advancements in the AI co-clinician research initiative, focusing on two key areas: augmenting clinicians and real-time multimodal capabilities in telemedical settings. For clinician augmentation, the AI demonstrated superior performance over leading evidence synthesis tools and other frontier AI systems in objective analysis of primary care queries and medication knowledge tests. In telemedical settings, the AI co-clinician, leveraging Gemini and Project Astra, was tested for its ability to use live audio and video to engage with patients, simulating telemedical calls. While expert physicians outperformed the AI overall, particularly in identifying 'red flags' and guiding physical examinations, the AI performed comparably to or exceeding primary care physicians in 68 of 140 assessed areas. The post also highlights the engineering of trust through a dual-agent architecture (Planner and Talker) and rigorous evidence verification and citation checking.
Advertisement