Google unveils MedLM, a family of healthcare-focused generative AI models

Google has introduced MedLM, a family of generative AI models fine-tuned for the medical industry. These models, based on Google's Med-PaLM 2, are designed to assist healthcare workers in completing tasks. MedLM is available to Google Cloud customers in the U.S. (in preview in certain other markets) through Vertex AI, Google's fully managed AI development platform. 

The family includes a larger model for complex tasks and a smaller, fine-tunable model suitable for scaling across tasks. Google is competing with rivals Microsoft and Amazon to establish a presence in the healthcare AI market, with the potential value reaching tens of billions of dollars by 2032.

One of the early users of MedLM is HCA Healthcare, a for-profit facility operator, which has piloted the models with physicians to draft patient notes at emergency department hospital sites. Another user, BenchSci, has integrated MedLM into its evidence engine for identifying, classifying, and ranking novel biomarkers. However, the adoption of AI in healthcare has been met with mixed success historically, with concerns about accuracy and potential risks. The World Health Organization has warned about the risks of using generative AI in healthcare, emphasising the need for caution to avoid errors, patient harm, and erosion of trust in AI.

While generative models like those in Google's MedLM family are more sophisticated, research has shown that they may not be highly accurate in answering healthcare-related questions. Concerns about potential misinformation, data leaks, and errors in critical tasks have led to calls for cautious adoption and thorough testing of these technologies in healthcare settings. Google emphasises its commitment to a safe and responsible use of generative AI in healthcare, acknowledging the need for caution and responsible implementation.