Google Launches Generative AI Medical Model MedLM
-
Google recently released the MedLM generative AI model, specifically designed for the healthcare sector. According to the company's blog post, the first two versions of this model are now available to Google Cloud customers in the U.S., provided through Google's Vertex AI platform, which uses machine learning workflows to guide users in training, evaluating, and deploying generative AI models.
The MedLM model is based on Med-PaLM2, Google's second-generation "medically tuned" AI tool, which is part of the large language models (LLMs). Google stated that Med-PaLM2 achieved an 85% accuracy rate on U.S. medical licensing exam questions this year, equivalent to the level of an "expert" physician. Compared to the first Med-PaLM model released last year, Med-PaLM2's score improved by 18%.
The difference between the two MedLM models mainly lies in their size: "The first MedLM model is larger and designed to handle complex tasks. The second is a medium-sized model that can be fine-tuned and is best suited for scaling across various tasks," Google explained in the blog post.
Google also plans to integrate Gemini-based models into the MedLM series in the coming months to further expand MedLM's artificial intelligence capabilities.
The newly launched AI tools are designed to serve various aspects of the healthcare industry, including hospitals, drug development, and patient-facing chatbots. In a blog post, Google highlighted that some organizations have already begun testing the MedLM suite, noting that many of them "are incorporating it into production solutions or expanding their testing scope."
In a pilot project with HCA Healthcare, MedLM technology was used to enhance Augmedix's technology, which aims to listen to conversations between clinicians and patients and automatically translate them into medical records. Google stated that combining MedLM's generative AI with Augmedix's natural language processing will help continuously improve the quality of generated records while also making it easier and more cost-effective to scale transcription technology across hospitals.
BenchSci added MedLM to its existing AI-driven R&D platform to enhance the ASCEND engine's ability to quickly screen large amounts of clinical data and identify links between certain diseases and biomarkers.
Accenture and Deloitte have also become early adopters of MedLM tools. Accenture aims to accelerate AI processing of clinical documents, registration forms, medical claims, etc., while Deloitte seeks to improve the efficiency of contact center agents in patient matching.
Google has announced that it will continue to develop and introduce new MedLM models in the healthcare sector to further strengthen its influence in the field of artificial intelligence.