Google is looking for a win in healthcare by trying to stand out in the competitive world of artificial intelligence.
The company rolled out several AI-related updates at a healthcare event last week including adding AI-generated medical information to its search function, launching an AI co-scientist to its Gemini large language model and creating more ways for developers to integrate medical records into its Google Health Connect application. Its Google Cloud division also struck a partnership with Quest Diagnostics to use generative AI for insights into patients’ lab data.
Related: Epic, Salesforce and the race to build healthcare AI agents
Google has a complicated history with healthcare including dismantling its health division in 2021. Also, questions have been raised about the effectiveness of its healthcare-specific AI models and whether they consistently deliver high-quality health information.
The company is working to build trust through its industry partnerships, said Dr. Karen DeSalvo, Google's chief health officer and a former health official during the Obama administration. "Healthcare moves at the speed of trust," she said.
The interview has been edited for length and clarity.
How do you build trust when building AI models for healthcare?
We don't run hospitals. We don't have our own research labs the way a scientist does at a university. We count on having that feedback loop with them. We’ve made this the art of the possible, can they use it in the real world? We work that iterative cycle with our partners. We integrate clinicians with real-world experience like myself into the development of our research and products. These are people caring for patients, who have worked for health systems in the U.S. They’ve done this work and they can share ways the model would be useful.
How are you differentiating yourselves in this competitive AI market?
We’ve been a company that is very heavily focused on the ecosystem, whether we're curating it like on YouTube or with healthcare, or building it like with Android and the cloud. It's very much in our ethos that we want to work with others. We feel excited that we offer a full stack for customers including the development of the models and the ability to expose them on our platforms like Cloud and Android. We can also showcase how those models might work on our Pixel devices.
Also, we’re a consumer-facing company. Most of our multibillion-dollar platform products are consumer facing, and so we've got decades of working with and understanding how to not only build global solutions, but localize them where it matters. As the world evolves into a more generative AI-driven experience in healthcare, it's going to get increasingly personalized. It's going to mean that every person is going to have more data and knowledge about themselves in their own hands, and going to expect more from the systems with whom they're interacting.
How ready are healthcare organizations to use AI beyond basic capabilities?
Healthcare moves at the speed of trust. The more we can work with the ecosystem, including physicians and nurses and others, to understand the technology, the more they'll begin to understand what it's capable of and how to how to manage it. We just announced we are pushing out a training initiative with the Digital Medicine Society to help healthcare providers better understand general purpose AI. How is it different than other AI they might have used? How can they start working with it? That’s how we’re trying to help build trust.
How many electronic health record companies are you working with to bring medical records into Google Health Connect?
We have been talking to a lot of U.S.-based partners and in other countries such as Japan, India and the U.K. One of the challenges is that sometimes the information is too deep and rich in EHR, sometimes people need a summary. So that's led us to talking also with health plans that keep aggregated lists of medications. Sometimes they're a better source of truth because they're paying for the medications. We've made that connection available for general availability, and we've already had some 60 partners around the world working on integrations. We've got some companies that are hopefully going to join now that we're further along in general availability. So stay tuned.
Google is a member of the Coalition for Health AI and participates in other self-regulation initiatives. Where does the company see its role in terms of the AI industry regulating itself?
We’re supportive of generative AI for regulated use cases. We've been working to make sure people understand the technology and where it's going. Some of our work with our products such as Fitbit and Pixel are regulated. In those cases, we will work with regulators and follow processes that need to be followed. In this new age of general purpose AI, it's important for people to really understand that these are probabilistic models. They are not deterministic. Assessing them requires a different set of tools in the toolbox. We need to be thoughtful about the ways we approach regulation. For us, it’s do no harm, safety first. I will say a lot of the work we do with consumers does not require regulation to help the most amount of people. Whether it’s information gathering or its sharing personal insights, we can still do a lot of good without getting into the regulated medical space.
