IoT news of the week for Nov. 27, 2020

Morse Micro has scored $13M for Wi-Fi HaLow chips: Morse Micro, whose COO was a guest on the podcast earlier this year, has secured $13 million in additional funding, bringing its Series A round to $30 million. Morse Micro will use the new funding to expand its product and technology development teams and allow it to continue investing in its Wi-Fi HaLow chips. Wi-Fi HaLow is a standard designed to reduce the power consumption of Wi-Fi and let it extend further into places like basements so as to make the radio standard more palatable for battery-operated IoT applications. I’m not sold, but apparently investors are. (Morse Micro)

Investments in infrastructure should be investments in IoT: As we replace our aging infrastructure, we should add connected sensors to the newly constructed roads or bridges so we can monitor their health and also derive additional insights, such as usage. For example, this story is all about sewage systems that have had sensors installed for the purpose of water conservation and management which are now being used (sometimes in combination with newly added sensors) to track the spread of COVID-19. Most sewage testing today is actually done by hand, with the samples shipped to labs to be tested for the novel coronavirus (at least in the U.S.). But if we could develop sensors that looked for specific RNA in real time, they could be added to a sewage sensing platform that would provide the relevant information. All of this is to say that the benefits of connecting our infrastructure are no longer theoretical, so we just need to figure out the best way to do it. (Financial Times)

Call OSHA, these sensors detect worker fatigue: While many of us slave away over a keyboard, plenty of workers are still engaged in jobs that require heavy lifting and physical effort. Warehouses, utilities, and construction all depend on people’s physical labor. And when those people get tired, they can get hurt. Now researchers at an Italian university have developed a package of sensors that wrap around the chest and waist to help detect when workers get fatigued and may need a break. The apparatus is bulky and needs work before being commercialized, but I’d love to see something like this be used as an unbiased, validated way to tell someone to take a break. A government or regulator could build out a program around such sensors that leaves management out of the equation when it comes to pushing workers past their limit. By connecting the sensors, it also creates an auditable trail that could detect employers who were cavalier about their workers’ safety. (IEEE Spectrum)

There’s a new Espressif chip in town: This is exciting news for makers and IoT developers. Espressif, the chip maker behind the popular WiFi-enabled Espressif ESP8266 and the ESP32 that’s inside half a dozen connected products in my home, has a new chip out called the ESP32-C3. This new chip uses a RISC-V core in place of the Tensilica core in the original ESP32, and as this article points out, borrows its pins from the ESP8266. I’m excited because it’s a RISC-V core designed for use outside of the internal chip teams at a big-name computing firm. (Hackaday)

Can AI reduce your bill of materials? In reading this story about how Amazon redesigned the fourth-generation Echo, going from the Pringles can to a bocce ball, there are some excellent tidbits. But what jumped out at me was that the first Amazon Echo device from 2014 had seven mics all arrayed on top of the “can” to separate it from the speaker and ensure Alexa could hear requests. The latest version has just two mics, in part because the algorithms that help Alexa understand speech have gotten better. Microphones inside a device aren’t a huge cost, but this story did lead me to wonder where else designers might cut components without compromising on quality as machine learning gets better and helps devices fill any gaps in their perception. (Fast Company)

Researchers want companies to stop selling AI models as science without proof: The beauty of science is that by forming a hypothesis and testing that hypothesis through a rigidly designed experiment, a scientist can determine if something is true. Others can then perform the same experiment and get the same results. But in many fields, scientists are discovering their experiments aren’t working that way, that even when people replicate their steps, they get different results. It’s like if sometimes you mixed blue and yellow paint and made red instead of green. According to this article, AI has this problem too, and it’s even more frustrating for scientists because the companies touting their science aren’t even explaining their inputs and models. And yet, that doesn’t stop companies from using AI for everything from the banal (identifying birds at a bird feeder) to the consequential (determining who makes bail). Scientists are fed up with the inability to replicate results touted by companies’ AI models, and are calling for more transparency from companies so researchers can see if their grandiose AI claims are merited. (MIT Technology Review)

We need to register our robots: My monthly column for IEEE Spectrum is all about why I think we need to develop a registry for the robots that will one day roam our streets and stores. We have a registry for drones, so click on through to understand why we need to be able to identify and understand other robots as well. (IEEE Spectrum)

Don’t buy no-name doorbells online: Security researchers trawled through Amazon and eBay to pick up video doorbells from off-brand manufacturers so they could test their security. Unsurprisingly, these devices are not what you’d want on your network. Some send your home SSD and password unencrypted over the air and store them in unencrypted servers online. That’s bad. And it’s why one of my go-to pieces of advice for people who ask how to buy secure IoT products is to look for name brands, and to Google those brands to see if they have had security issues and how they have handled those issues. Personally, I think Amazon and other retailers have a big role to play in ensuring consumers have the most secure devices as well. They shouldn’t stock products with poor security practices. (Dark Reading)

Do you need an overview on securing edge devices at the chip level? If yes, then this article is an excellent resource for understanding the difference between a root of trust and a secure enclave. From there, it talks about silicon-level efforts to stop attacks during the manufacturing process. It also explains why device-level security is important, but needs to be part of a systemic approach to security. You should have a sense of all these things if you’re responsible for an IoT project, so take a look. (EETimes)