Federated AI Still Has a Central Coordinator Problem

Federated Learning Isn’t Decentralized AI (Yet)

Here's the problem federated learning is trying to solve: Most AI training starts with pooling data. That fails wherever the data is sensitive or legally stuck. In 2023, Samsung reportedly restricted employee use of ChatGPT after engineers pasted internal source code into it. In late 2024, Italy’s data protection authority fined OpenAI €15M over how personal data was handled in training. Hospitals hold exactly the data that could improve diagnostic models, but they can’t simply merge patient records.

Federated learning is the standard answer: instead of moving data to the model, move the model to the data. Each site trains locally and sends only model updates to a coordinator, which aggregates them. It’s not theoretical. Google has used it for on-device keyboard prediction in Gboard, and a multi-hospital study (EXAM) trained a COVID-19 oxygen-needs model across roughly 20 institutions without sharing patient...

Copyright of this story solely belongs to hackernoon.com. To see the full text click HERE