Google, Microsoft and startups test out generative AI in healthcare

It's been five months since Microsoft-backed OpenAI released its generative large language model ChatGPT, followed by GPT-4 in March.

Now, tech giants and startups are off to the races to test out the potential for LLMs and generative AI tools in medicine, clinical settings and research.

This week, Google announced it's releasing a version of its medical LLM, called Med-PaLM 2, to a limited group of users. "It will be available in coming weeks to a select group of Google Cloud customers for limited testing, to explore use cases and share feedback as we investigate safe, responsible, and meaningful ways to use this technology," Google Cloud leaders Aashima Gupta and Amy Waldron wrote in a blog post.

The tech giant continues to invest in medical large language model research. LLMs are AI tools that demonstrate capabilities in language understanding and generation. Google developed Med-PaLM 2 late last year as a version of PaLM tuned for the medical domain to more accurately and safely answer medical questions, Google executives said during the company's annual The Check Up event last month.

Google claims that Med-PaLM 2 was the first LLM to perform at an “expert” test-taker level performance on the MedQA data set of US Medical Licensing Examination-style questions, reaching 85%-plus accuracy, and it was the first AI system to reach a passing score on the MedMCQA data set comprising Indian AIIMS and NEET medical examination questions, scoring 72.3%.

"We’re looking forward to working with our customers to understand how Med-PaLM 2 might be used to facilitate rich, informative discussions, answer complex medical questions, and find insights in complicated and unstructured medical texts. They might also explore its utility to help draft short- and long-form responses and summarize documentation and insights from internal data sets and bodies of scientific knowledge," Gupta, global director of healthcare strategy and solutions, and Waldron, global director of health plan strategy and solutions, wrote in the blog post.

Google has been working with clinicians and non-clinicians to assess Med-PaLM and Med-PaLM 2 against multiple criteria including scientific consensus, medical reasoning, knowledge recall, bias and likelihood of possible harm, executives said.

This week, Google also unveiled its new AI-enabled tools designed to streamline processes for health insurance prior authorization and claims processing. The tech converts unstructured data into structured data that help experts make faster decisions and improve access to timely patient care and is available to Google Cloud customers, the company said.

Blue Shield of California is one early adopter of the Claims Acceleration Suite along with international healthcare company Bupa.

"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," Lisa Davis, senior vice president and chief information officer at Blue Shield of California, said in a statement.

Google's announcement this week follows closely on the heels of Microsoft rolling out new generative AI tools embedded in its cloud computing capabilities. The company released a new Azure Health Bot template to enable healthcare organizations to experiment with the integration of Azure OpenAI Service into their chatbots.

Currently, Microsoft is previewing the new health bot template for internal testing and evaluation purposes only.

"There's been so much discussion around large language models and generative AI and now having the Azure health bot integrated with the OpenAI service so that you can actually answer not only those questions that you had previously defined, but it would allow you to be able to answer things in a way that you're searching for using validated sources," said David Rhew, M.D., global chief medical officer and vice president of healthcare at Microsoft. "These [tools] are all building blocks to help us achieve better health outcomes and better coordination of care."

In February, Doximity, a digital platform for medical professionals, rolled out a beta version of a ChatGPT tool for doctors that helps streamline some of their time-consuming administrative tasks, such as drafting and faxing preauthorization and appeal letters to insurers.

The open beta site, called, is an integration with ChatGPT that works with Doximity’s free fax service, said Jeffrey Tangney, Doximity co-founder and CEO, during the company's fiscal 2023 third-quarter earnings call Thursday.

As companies race to create their own AI models, the real opportunity is in pairing these technologies with companies to create trusted models to specialize in the healthcare industry, according to executives at Pittsburgh-based startup Abridge, which provides AI-powered medical transcription services and is behind what is at the moment one of the largest deployments of generative AI in healthcare. The University of Kansas Health System has rolled out Abridge's tools to more than 1,500 physicians across the system’s more than 140 locations. Abridge's technology listens to patient-doctor visits and uses smart tech to summarize the most important parts of the conversation for clinicians and patients.

The University of Pittsburgh Medical Center also plans to roll out the tech to thousands of physicians to streamline medical note-taking.

While other tools are available to help with documentation, such as medical scribes, they are becoming increasingly expensive and unreliable due to turnover, Greg Ator, M.D., a practicing ENT physician and chief medical informatics officer at the University of Kansas Health System, told Fierce Healthcare.

Abridge’s technology identifies more than 90% of the key points from provider-patient conversations and quickly generates summaries in various formats meant for clinicians or patients, the company said.

Researchers and clinicians continue to be cautious about the use of generative AI in healthcare given the technology’s tendency to “hallucinate,” or invent a response when it doesn’t have sufficient information.

According to a recent survey by generative AI tech company Huma.AI, medical affairs leaders see generative AI playing a  big role at life sciences companies in the next two years.

The majority (86%) of the medical affairs leaders when asked whether generative AI can be applied at life science companies answered, "Yes" or "eventually Yes." Generative AI holds the potential to transform medical affairs activities, helping teams provide more effective and targeted engagement with healthcare professionals and ensure products are used safely and effectively by HCPs and patients, among other uses. 

“Generative AI could enable more efficient and effective analysis of vast amount of unstructured data such as internal data and publications, strategy development, and decision-making,” said Lana Feng, Ph.D., CEO of Huma.AI. “However, it is important to note that generative AI should not be seen as a replacement for human expertise and judgment, but rather as a tool to augment and supercharge human decision-making.”