TinyML is about to get really big, or at least that’s what a startup thinks, as we explain on this week’s podcast. Useful Sensors is the company that’s making inexpensive, low-powered edge sensors in a way that protects privacy. We discuss why we agree with that approach. Next up are our thoughts on why 5G really hasn’t taken the IoT market by storm yet. You’ll want to hear our reasons for this because there are several. We then turn to Apple, as the company is reportedly working on an iPad smart display of sorts, similar to the recently announced Google Pixel Tablet. Kevin then points out something important about the newest Apple TV 4K devices if you plan to have a Thread network for Matter devices in your house. And speaking of Matter, there’s a new USB dongle for HomeAssistant: It adds multiple radios for both Matter and Zigbee support. If you’d rather go with an integrated solution, we can point you to Aeotec as we share news of its SmartThings hub upgrade for Matter. Health data and algorithms also make the show this week as there’s a unique project to track which algorithms are better than others. Lastly, we discuss Verizon’s free new service for its internet customers: Verizon routers can now detect physical disruptions in your network, indicating the movement of people. Before closing out the news portion of the show, we answer a listener’s question about people controlling smart home devices in vacation homes.
Our guest this week is Pete Warden, CEO of Useful Sensors, a company that’s bundling a sensor with predetermined machine learning algorithms for recognizing people, faces, gestures, and more. Warden explains the challenges of TinyML; the act of embedding machine learning algorithms on constrained, power-sipping devices; and how he hopes Useful Sensors can help companies that build devices figure out compelling uses for the technology. TinlyML has a huge amount of promise for the IoT, but it’s hard to find use cases outside of the ubiquitous wake-word detection. By offering a $10 sensor that can provide person and face detection to makers, Warden hopes to jumpstart new ideas for TinyML. We might see those in future appliances, televisions, toys, and more. We also talk about how he’s thinking about respecting consumer privacy and what it will take to make people feel comfortable in a world with millions of tiny cameras, microphones, and other sensors embedded in everyday objects. Enjoy the show.
I can certainly see TinyML being a starter of sorts for the hobbyist line of IoT functions. And I can certainly see ML for full feature work on bigger systems. But inside actual hardware for that purpose? Not now.