How does federated learning let AI train on your data without ever seeing it?

The answer

Federated learning, pioneered by Google in 2017, is a distributed machine learning approach where models are trained across many devices (phones, hospitals, banks) without centralizing the raw data. Each device trains a local model on its own data and sends only the model updates (gradients) to a central server, which aggregates them into an improved global model. Google uses it for Gboard's next-word prediction, training across millions of Android devices. Apple adopted it for Siri improvements and health features. In healthcare, federated learning has enabled multi-hospital cancer detection models that would be impossible under traditional data-sharing regulations. Challenges remain, including communication overhead and vulnerability to gradient inversion attacks that can partially reconstruct training data.