Google deepens AI push into healthcare with Fitbit, screenings

Alphabet Inc.'s Google announced a slew of initiatives to deploy its artificial intelligence models in the healthcare industry, including a tool that will help Fitbit users glean insights from their wearable devices and a partnership to improve screenings for cancer and disease in India.

At its annual health event in New York City, the company said Tuesday that teams at Google Research and Fitbit, which it owns, were developing a new AI feature that will draw data from the wristbands to coach users on their personal health. Powered by Google's most advanced AI model, Gemini, the tool could, for example, assess how workouts affect the quality of a person's sleep.

The tech giant also said it will work with Apollo Radiology International to deliver AI-powered screenings for tuberculosis, lung cancer and breast cancer in India. Once Apollo has secured regulatory approval, the effort will deliver 3 million free scans over the next 10 years, Google said. The company said it's also improving health information in Google search and YouTube, its video site.

Google has been trying to revolutionise healthcare for years, with varied success. It announced plans to buy Fitbit in 2019, but the deal came under heavy antitrust scrutiny amid concerns about what Google would do with the vast user data it collected. Last year, Google unveiled Med-PaLM, an AI model that has capably fielded medical questions.

The latest initiatives are part of Google's pledge to ensure that high quality health information is being delivered across all its products used by consumers and companies, Karen DeSalvo, Google's chief health officer, said in an interview.

The company considers "what are the big health challenges" in the world and how its technology "can make a difference, whether it's for the individual or for populations," she said.

Mountain View, California-based Google has been striving to integrate its AI work, including its large language models, into medical research. Large language models are massive systems that ingest enormous volumes of digital text "- from news articles, social media posts and other sources "- and use that written material to train software that predicts and generates content when given a prompt or query.

On Tuesday, Google said that it had begun to explore fine-tuning its Gemini AI model for medical applications. The personal health tool being developed in collaboration with Fitbit will also be powered by Gemini, Google said. The company didn't disclose a timetable for when the feature will be released to the public.

Greg Corrado, a senior research scientist at Google, said the surge of interest in AI has set the stage for important discussions between Google and medical professionals about how the technology can best be deployed in healthcare.

"Its rise to prominence is the opening of a door," Corrado said of AI. "It helps people understand that there's a new realm of possibilities." Google needs to "engage in a dialogue with healthcare providers, with doctors, with patients, with insurers to understand how the technology can be useful. It's that conversation that I think is going to drive change over the next decade."