Google announced MedLM, a suite of health-care-specific artificial intelligence (AI) models, designed to support clinicians and researchers in complex studies and summarize doctor-patient interactions. MedLM, part of Google’s broader initiative to monetize health-care AI tools, showcases the company’s commitment to competing in the market alongside giants like Amazon and Microsoft. The suite, which includes both large and medium-sized AI models built on Med-PaLM 2, is now available to eligible Google Cloud customers in the U.S.
MedLM’s larger model excels in tasks requiring deep knowledge, such as conducting studies with extensive patient data. The medium-sized model, on the other hand, is optimized for more specific or real-time functions like summarizing doctor-patient interactions. Google also plans to introduce health-care versions of Gemini, its most capable AI model, to MedLM, enhancing its capabilities further.
Companies like HCA Healthcare have been testing Google’s AI technology, using MedLM to automate documentation in emergency rooms and develop tools for nurse handoffs, significantly reducing the time spent on clerical tasks. However, challenges remain, such as managing incorrect AI outputs and token limits.
Google’s MedLM has been tested extensively with customers, leading to its general availability. The company emphasizes the importance of controlled health-care settings for testing and evaluating new models like Gemini before broader rollouts. Partners and clients like BenchSci and Deloitte are already exploring MedLM’s potential in drug discovery and patient care, albeit with careful validation and continuous feedback for improvement.
This development marks a significant step in integrating AI into health-care workflows, promising to enhance efficiency and potentially transform patient care. However, the careful implementation and ongoing evaluation of these AI models remain crucial to ensure their beneficial impact on the health-care industry.
Source: CNBC
Image by PublicDomainPictures from Pixabay
