Google SensorFM turns wearable data into a foundation model for health signals

Google Research introduced SensorFM, a foundation model trained on large-scale wearable sensor data from Fitbit and Pixel Watch users. The Decoder reports that the model beat existing benchmarks on 34 of 35 health and behavioral tasks.

The work points toward AI systems that can reason over messy real-world signals rather than only text, images, or audio. Google has not announced a product integration, but SensorFM could eventually support more adaptive health coaching and behavior-aware assistants.