Google announced it’s building a new large language model that will give health and fitness tips on Tuesday, part of a broader push across the company to develop AI health algorithms and robot doctor tools.
Fitbit (a subsidiary of Google) is partnering with Google Research on the project, which the company calls “Personal Health Large Language Model.” In a press release, Google said it’s fine-tuning the algorithm to give personalized advice based on the data harvested by Fitbit and Pixel devices. Thanks to the latest generation of AI chatbot tech, the new Fitbit health coaches will be able to use a language-based interface to send users “actionable” messages and guidance.
“For example, this model may be able to analyze variations in your sleep patterns and sleep quality, and then suggest recommendations on how you might change the intensity of your workout based on those insights,” said Yossi Matias, Google’s Global Lead of AI for Health, Climate, and Education in a blog post.
According to Google, this new chatty health bot is already performing well on practice tests based on sleep medicine certification exams. The new tool will run on the company’s Gemini AI model. Google says it’s collecting anonymous data from real, human case studies gathered from accredited coaches and wellness experts to train up the AI. The company didn’t share a release date for the project, but promised updates “soon.”
The Fitbit AI is just one of a slew of health-related AI projects at Google. Last week, Google started rolling out an AI tool to some of its Cloud customers called MedLM for Chest X-rays, which helps doctors screen for operations, radiology tests, and diagnosis. The company also announced that a specially-tailored version of its Gemini model scores 91.1% on benchmark tests based on the U.S. Medical Licensing Exam. Earlier this year, Google researchers published a study demonstrating that their AI has a better bedside manner than human doctors and makes better diagnoses.
Google also announced it’s partnering with a company called Apollo Radiology International in India to provide AI-powered screenings for TB, breast cancer, and lung cancer over the next decade, and it’s partnered with Stanford Medicine on a new dataset of photos of skin conditions to help dermatologists train new skin-scanning tools for diagnosis.