IoT news of the week for Nov. 29, 2019

Cellular IoT will reach 5 billion connections by 2025: — Ericsson has put out its November 2019 Mobility report with fun data on 5G, smartphone data consumption, and the IoT. For the IoT, the big news is that we’re about to hit an inflection point in IoT-specific cellular connections using low-power and broadband-capable connections. By the end of 2025, NB-IoT and Cat-M are projected to account for 52% of all cellular IoT connections. The report also includes data on the adoption of 5G based on the types of devices that will be available, and the shocking stat that by 2025 smartphone data consumption will reach 24 GB per month. Excuse me, I need to go adjust my data plan. (Ericsson)

This is a fun gift and where I thought the IoT would go: A Boston store is selling connected maps that show a real-time version of the subway cars on the city’s transit system. The creation was designed by two gentlemen at a Somerville, Ma., hackerspace, but I love that they will soon sell a larger version. Who wouldn’t want this, or a localized map like this that pulls in real-time data from a metropolitan transit system to act as both a talking point and as a hint of your commute ahead? I was thrilled when I saw it, because when I started writing about IoT, I really thought these sorts of highly personalized devices that users could customize by pulling in public APIs would be everywhere. But the process of building the physical hardware was still pretty hard, and the public APIs needed to power something like a living transit map were lacking. My hoped-for future hasn’t arrived, but this gift does give me hope. (The Boston Globe)

Wall Street hates dedicated consumer hardware companies: This week, we found out that Facebook was the other suitor for Fitbit, which Google has agreed to purchase for $2.1 billion. Not only does that tell us that this deal was about the data, but the proxy statement filed as part of the M&A docs also contained a chart showing how poorly the market values dedicated hardware companies. The chart makes for depressing reading for Sonos, Arlo, and the many companies building smart gear that may have hoped an IPO would provide a promising exit. No wonder Peleton wanted to be a media business. (Fortune)

Asset tracking is the killer app for LPWAN: Based on this excellent profile of Sigfox, a proprietary Low-Power Wide-Area Network, asset tracking is the big draw for it and other such LPWANs. Asset tracking accounts for 80% of Sigfox’s revenue and by the end of the year, it expects to have 16 million connected devices. That is nothing compared to the 1 billion devices the CEO expects to have on the network in four years. But to get there, Sigfox needs cheaper modules and energy-harvesting technology to avoid having to change hundreds of millions of batteries. Sigfox modules are currently priced at 20 cents each. Meanwhile, it’s experimenting with silliness such as using fruit to power those modules. The full article is worth a read. (EETimes)

Qualcomm and Siemens create 5G factory network: Qualcomm and Siemens have set up a 5G testbed in the Siemens Automotive Test Center in Nuremberg, Germany, with a private 5G network running in the 3.7-3.8GHz band. This is really 5G in the sense that it is using both the cellular radio standard and the high-frequency spectrum bands that will allow for huge amounts of data to be transferred. At lower frequency ranges, the blocks of spectrum allocated for data are too small for gigabit capacity. You need channels that are 100 megahertz wide for that sort of capacity. Of course, the higher the wavelength, the tougher the transmission. At this range, people, machines, and more will get in the way of the radio waves and cause attenuation. The test is also an example of a private network as opposed to one operated by a mobile operator. I expect to follow this initiative closely because I really think the use of small, locally-owned spectrum for private 5G networks is going to lead to a big shift in connectivity. (Qualcomm)

Alexa gets an attitude: We can’t help but anthropomorphize our digital assistants, but now Amazon wants to make it even easier. Amazon will now let developers choose between a “happy/excited” or a “disappointed/empathetic” tone for Alexa when answering user questions. Doing so will help make Alexa a bit more personable, which will help Amazon position Alexa more as a helper and less as a robotic extension of Amazon’s commerce engine. (The Verge)

I wrote about creating a devops-style process for machine learning: This story came together after a visit to Microsoft, where someone mentioned that companies really need to document how they are building and training algorithms. The belief is that if algorithms are recipes that will soon underlie crucial parts of the business, it’s important to know how the recipe was created. And it’s a good idea, in case the company’s head data scientist leaves or if auditors come calling with questions. I was surprised that people weren’t already doing this, so I figured I’d point it out in my monthly IEEE Spectrum column. (IEEE Spectrum)