IoT news of the week for Oct. 14, 2022

Microsoft and Mercedes-Benz built up a relationship in the cloud: Microsoft wants to be the IT side of any IT-OT convergences, and a new deal with Mercedes-Benz shows how it has built infrastructure to bring plant data into IT environments for analysis. Mercedes-Benz, meanwhile, has built its MO360 Data Platform to connect 30 passenger car manufacturing plants to Microsoft Cloud to help meet sustainability and production goals. (Microsoft)

ADT closes on its $1.2B investment from State Farm: Security firm ADT has closed on the previously announced $1.2 billion equity investment from State Farm. State Farm has also given $100 million of a planned $300 million in cash for R&D into new safety and insurance products created by ADT. As we said when the deal was initially announced, the push by insurance firms is a long time coming, and the decision to do so in conjunction with a security and smart home company makes sense. Both products are about understanding and then preventing or mitigating risks. (ADT)

The right to longevity in connected devices? With Google killing its Stadia cloud gaming service, there are renewed calls for hardware to retain some usefulness after that hardware’s core cloud service is shut down. Google’s Stadia controllers are beloved by gamers as they connect directly to the cloud as opposed to a computer, reducing latency. And those gamers, who spent $70 on the devices, would like to keep using them, so there are again calls for rules that require source code for gear to be put in escrow or released upon the death of a device. We have long asked for expiration dates on connected devices and escrow for code, and for startups to think about what might happen should their business fail and strand a bunch of working hardware. The UK has rules that call for expiration dates on connected devices and mandates security updates for up to three years after a product is sold, so we’re slowly getting to a place where regulators realize that connected devices don’t have to die when the cloud does — and that such deaths contribute to e-waste. (Petafloptimism)

What do eSIMs mean for the IoT? I’ve been waiting for eSIMs to reach the popular culture for a decade, and with Apple’s latest iPhone, they have. This article does a good job of explaining why having over-the-air-provisioned SIMs that aren’t tied to a specific operator will be good for the IoT. Reasons include simplicity, in that device makers can have eSIMs on devices and won’t have to physically manhandle them to change carriers or manage locations. This also gives them flexibility when it comes to pricing connectivity. Plus eSIMs have required security features that will improve the overall security of the connection with the device. (Times of India)

Two fundings for IoT companies: Smart light switch maker Deako raised $13 million from what the CEO described as “industry investors,” brining its total funding to almost $50 million. Deako’s switches are modular and sold to home builders for installation into their new builds. Customers include DR Horton and KB Homes. The other funding is $21 million for SecuriThings, a company that sells a security software to enterprises. Its software identifies and monitors connected devices installed on a network and can be programmed to automatically perform tasks such as update firmware and manage certificates devices. (Geekwire, TechCrunch)

Will Roku buy Wyze? This charming bit of speculation is at the bottom of a Protocol newsletter about Roku’s smart home devices. And because the story is by my guest on this week’s show (and friend) Janko Roettgers, I’m going to give the matter a lot of thought. He’s pretty smart. Be sure to check it out after the Meta story. (Protocol)

Peachtree Corners gets smarter parking with 5G: Peachtree Corners, Georgia, which is a smart city project where all kinds of companies test out new connected infrastructure concepts, has a new parking solution that uses 5G from T Mobile, a sensor package from Bosch, and Cradlepoint’s wireless infrastructure gear. The new parking product helps Peachtree Corners’ visitors find available parking and charging stations for electric vehicles, and provides information about real-time traffic to city officials. (Enterprise IoT Insights)

Two edge ML players team up for local computer vision: Imagimob, a platform provider for tiny ML algorithms and Syntiant, a company that builds AI chips for local machine learning, have signed a partnership to create a way to easily get Tiny ML models built in Imagimob onto Syntiant’s chip with a click of a button. The combined Imagimob-Syntiant solution supports several applications such as sound event detection, keyword spotting, fall detection, anomaly detection, gesture detection and more. For developers familiar with Imagimob’s platform this gets them into the real world quickly. And for those already working with Syntiant’s chips, it’s a way to steer them toward Imagimob’s platform. (Imagimob)

To automate your supply chains, consider an autonomous agent: For companies that want to create more visible and predictable supply chains, one idea is to integrate data from all of their suppliers into their own systems. But that integration is hard for smaller suppliers and can offer little benefit to them outside of that one relationship. This is where the idea of some kind of software agent comes in. The agent gets installed at the supplier and handles the relevant data, sending it to the buyer and keeping a digital eye on potential problems. This is a pretty cool idea and sounds like less work than creating a digital twin, so check out this concept as explained in the linked article. (The Manufacturer)

MIT researchers have an idea to train ML models for the edge: I write all the time about machine learning at the edge and Tiny ML, but for the most part when I cover edge ML I’m writing about inference, where a pre-trained algorithm handles incoming data, running that new data against the algorithm to see if it matches. What’s more exciting is training an algorithm. Today that’s done in the cloud, but if it could happen at the edge, it could open up a world of personalized ML on devices. Devices could get smarter without ever talking to the cloud and sharing your data. MIT researchers think they have found a way to train machine learning models using less than a quarter of a megabyte of memory, which gets much closer to not just robust edge devices, but even microcontrollers. (MIT News)