Google Shares Progress on Generative AI Initiatives

This week, Google provided progress updates on multiple of the company’s healthcare artificial intelligence (AI) initiatives, including an announcement that its medical large language model (LLM), Med-PaLM 2, would become available to select customers for limited testing. 

Med-PaLM 2 is a version of Google’s original LLM, known as the Pathways Language Model (PaLM). PaLM was introduced last year, and Med-PaLM, a version tuned for the medical domain, was developed shortly after.  

Med-PaLM achieved a passing score on USMLE-style multiple choice and open-ended questions, and Med-PaLM 2 recently scored 85 percent on similar questions. However, these tests highlighted gaps in the model’s ability to effectively answer medical questions, which Google indicated it would address through additional research and testing.  

Some of this testing is about to be underway, as this week’s announcement stated that select customers will be granted limited access to Med-PaLM 2 in the coming weeks for testing. During testing, users will share feedback on the model and explore its potential use cases, such as answering complex medical questions, summarizing documentation, and generating insights from unstructured data. 

These efforts are part of Google’s ongoing work in generative AI technologies, which can find relationships within large sets of data and use these to generate new data in the form of text, images, audio, videos, or synthetic data.  

In addition, Google also announced its Claims Acceleration Suite, an AI-driven solution designed to streamline prior authorization and claims processing for healthcare providers and payers. The tool leverages Google’s Claims Data Activator to help reduce administrative burden by converting unstructured data into structured data to drive efficiency in claims decisions. 

Blue Shield of California is an early adopter of the solution and aims to use it to improve provider and member experience. 

"Streamlining the prior authorization process is one important part of Blue Shield of California's journey to enable a more real-time exchange of data and provide more retail-like experiences. We want to help ease the administrative burden on our healthcare providers so they have more time to deliver the best care possible," said Lisa Davis, senior vice president and chief information officer at Blue Shield of California, in the press release. "Leveraging Google Cloud technologies and artificial intelligence, we are working to ensure our members get timely access to clinically necessary care and services. What sets Google Cloud apart is their commitment not only to technical capabilities but also to connecting the healthcare ecosystem through interoperability and using open standards." 

These announcements come a month after Google shared health AI research updates and new partnerships to apply AI to cancer, tuberculosis, and maternal health at its annual health event, The Check Up. 

The first of these was a formal agreement with Mayo Clinic to investigate whether AI can help streamline cancer radiotherapy. 

A second collaboration with Jacaranda Health, a Kenya-based nonprofit focused on improving health outcomes for mothers and babies in government hospitals, to explore how AI-assisted ultrasound may improve image interpretation, boosting maternal and infant outcomes. 

Other partnerships will focus on early breast cancer detection and AI-powered tuberculosis screening.