In a significant development within the generative AI landscape, Google has launched MedLM, a series of foundation models specifically designed for healthcare applications. Available through Google Cloud, this move add-up to the company's ongoing endeavour to integrate advanced AI technologies into the healthcare sector.
Following the introduction of Med-PaLM, a model developed to address medical queries, Google said it has now expanded its healthcare AI offerings with MedLM. This initiative comprises two models, each based on the framework of Med-PaLM 2, and tailored to meet varying healthcare needs.
The larger model is engineered for more complex tasks, while the medium-sized model focuses on scalability and customisation. Both models have been made available to U.S. customers via Google's Vertex AI platform.
MedLM's Future Prospects
Google's roadmap for MedLM includes the integration of Gemini-based models, aiming to further enhance its range of AI tools for healthcare. Gemini, representing Google's latest large language model, positions itself alongside other major players such as OpenAI and Microsoft's GPT-4.
According to Google, various companies are currently engaging in the testing and implementation of MedLM, either incorporating it into their existing systems or extending their trial phases.
MedLM in Healthcare Applications
In the practical realm, HCA Healthcare has been utilising MedLM in its emergency departments to assist with medical documentation. Augmedix, a company specialising in ambient medical documentation, has transitioned from using Google Cloud's Med-PaLM 2 to integrating MedLM into its system.
Augmedix's use of Google Cloud's Vertex AI platform for model refinement is aimed at improving the accuracy and efficiency of medical documentation.
BenchSci is another example of MedLM's application, as it incorporates the model into its AI-driven drug discovery processes. Deloitte and Accenture are also employing MedLM to enhance various facets of patient care. These collaborations reflect the wide-ranging applicability of MedLM in the healthcare industry.
The adoption of MedLM by organisations like Augmedix highlights the increasing demand for specialized AI solutions in healthcare. Augmedix's strategy to integrate MedLM is intended to improve both the quality and speed of its services, which currently cover a broad spectrum of medical specialties and involve substantial medical documentation.
The use of MedLM in BenchSci's ASCEND platform is anticipated to boost the efficiency and effectiveness of pre-clinical research in pharmaceutical development.
Future of Medical Generative AI
Google said it is excited by notable progress within a year in medical AI. The company initially developed a language model (LLM) that achieved a passing score of over 60% on U.S. medical-licensing-exam-style questions, a milestone reported in the journal Nature.
Building on this success, the model was further refined to reach an expert level score of 86.5%. Google has also begun implementing this advanced AI technology in real-world healthcare settings, adopting a careful and methodical approach to its application, said the company.