番茄社区

Welcome to MedAI Roundup, highlighting the latest news and research in healthcare-related artificial intelligence each month.

Former FDA Director Scott Gottlieb, MD, and colleagues pitted several chatbots against each other in taking the U.S. Medical Licensing Examination (USMLE) Step 3. . (CNBC)

Physicians are increasingly turning to generative AI tools to . (New York Times)

AI tools may be able to help researchers better "" to their data. (Nature)

Generative AI responses to patient messages were more empathetic than physician responses, but they were also wonkier, which may pose a challenge for patients with lower health or English literacy, according to a study published in .

GPT-4 Vision -- the multimodal version of OpenAI's latest model -- outperformed physicians on a multi-choice image-based quiz designed to assess diagnostic ability (81.6% vs 77.8%), but it also provided flawed explanations for 35.5% of its answers, according to research in .

Large language models (LLMs) built specifically with evidence-based medicine in mind bested general LLMs on answering clinical questions, according to a study conducted by Atropos Health and .

An article in warned about the potential risk of LLMs corrupting training data with their own previously generated data in a downward feed-loop, which could lead to "model collapse."

On the business side, the Mayo Clinic has partnered with Abridge and Epic to develop a generative AI ambient documentation workflow .

Microsoft is partnering with Mass General Brigham and the University of Wisconsin-Madison to develop AI-powered tools for .

Suki announced a collaboration with several Federally Qualified Health Centers and Community Health Centers aimed at addressing in underserved areas.

Commure , a developer of ambient AI clinical documentation tools. (Fierce Healthcare)

Comment