Google is testing AI that spots high blood pressure from smartwatch data
High blood pressure often has no symptoms, so lots of people have it and don’t know until someone checks. Researchers at Google, Beth Israel Deaconess Medical Center and Harvard tried another way to catch it: an AI model that reads what a smartwatch already records. They published the study in Nature on November 5.
The watch doesn’t measure blood pressure, and it doesn’t give you a number. Its sensors track your pulse and how your wrist moves. After several days of that data, plus your height, the AI estimates whether your blood pressure is likely to be high most of the time.
The researchers tested it on 196 adults at five US sites. None had known heart disease or took blood pressure medication. To check the watch’s answers, each person also wore a cuff that took readings over a full day.
The watch picked up about two out of three people with high blood pressure. A first check at a doctor’s office caught just over half. That gap was small enough that it could be down to chance, so the study doesn’t show the watch beats the office check. The watch also missed about one in three cases. Both approaches wrongly flagged about one in ten people whose blood pressure wasn’t high.
It’s a small, early study, and the watch isn’t diagnosing anyone. What it suggests is that a device many people already wear could one day give them a reason to get a proper blood pressure test, not replace that test.
Sources: Google Research blog · Nature press release
Robots that restock supermarket shelves overnight
Theseus Robotics, a startup based in Switzerland, says two of its robots are now working in supermarkets. At night they restock shelves from the back room, and during the day they pick items for online grocery orders. The company announced the deployment on November 7, and clips on its launch page show the robots in a store in Zürich.
These aren’t walking humanoids. They’re wheeled machines with two arms that Theseus buys from other manufacturers. What the company builds is the software. It has to recognize each product, work out how to grab it and put it in the right spot, using the store’s delivery list and shelf layout. A yogurt cup and a case of water need completely different grips, and Theseus says its robots handle both, along with thousands of other items.
Theseus says the robots can restock up to 400 items an hour and pick up to 40 items an hour for online orders. Those figures come from two robots in two stores, and the company hasn’t shared an average or any independent measurement.
People can also operate the robots remotely. Theseus says this lets it get robots into stores quickly and collect the data it needs to make them more independent. It hasn’t said how much of the work the robots currently do on their own.
The company has pilots planned for up to 10 more robots in 2027 and is looking for grocery, fashion and drugstore chains in the US, Switzerland and Germany to take part. When those pilots run, the number worth watching is how much shelving the robots do without a person at the controls.
Sources: Theseus Robotics launch post · Company page