AI Runs Your Smart Devices — But It Can't Hear Them, Study Finds

Every time you ask a smart speaker a question, or your car pulls up live traffic maps, computers make a split-second decision about where the work should happen. On the device itself? On a nearby server (engineers call this "the edge" — think a box in your neighbourhood rather than a warehouse across the country)? Or in a giant data centre far away? Increasingly, AI is being handed that juggling job. Researchers from Italy reviewed how this works and built a scorecard to compare the systems doing it.

They applied that scorecard to six real, recently built systems. The surprise: not one of them lets the AI both give the orders and fully hear back how things turned out at every level. In plain terms, the AI boss can assign work but never gets complete reports from the shop floor. The authors call this a missing "feedback loop" — the same thing that makes a manager ineffective when nobody tells them what actually happened.

Why should you care? Because these decisions show up in your daily life. Bad choices mean your video call stutters, your smartwatch battery dies early, your home hub delays, and data centres burn electricity that ends up on your bill. When AI can't hear from all layers, it keeps making the same mistakes instead of learning. The paper argues that fixing this one gap could make the whole system smarter without upgrading a single device.

The catch: this is a map, not a machine. The researchers didn't build or test a new system, so there's no proof yet that fixing the feedback gap actually speeds anything up. It's a peer-reviewed conference paper, meaning other experts checked the reasoning, not the results. A practical fix could be years away. And there's a quiet trade-off — giving AI more visibility across layers means it sees more of your data.