GM is simplifying its chip supply chain: I think this is a smart move on GM’s part and also an indication that as we put computers into more places, we need to consolidate platforms and simplify the underlying hardware and software. GM will co-develop chips with its suppliers as well as shrink the number of chip families it uses in its vehicles to three. It’s unclear exactly what those families are, but it’s a good plan. By reducing the number of chips and suppliers it works with, GM can buy chips in bulk and guarantee supplies better than a manufacturer that is buying fewer chips (say for one specific type of car). Doing so will also help GM pinpoint demand a bit better as it spreads the chips across better-selling vehicles and those that are in lower demand. All of this will help chip suppliers feel confident allocating production for GM’s parts. (Automotive News)
Ford is doing…something to address the chip shortage: While GM was a bit light on the specifics of how it plans to address the chip shortage with regards to the types of chip consolidation it will enact, it managed to provide some details, such as its overall strategy and the names of its chip partners. Ford, which is also trying to soothe the market with its plans to address the chip shortage, was much less clear. It has signed some kind of partnership with chip manufacturing firm Global Foundries, but we’re not sure what the partnership entails or how it will help. (Ars Technica)
No-code AI is becoming a thing: This week, Kevin showcased Clevr, a service that makes it easy to input images or text to train an AI model. It’s part of a wave of low-code or no-code options out there designed to make machine learning more accessible for people who have an idea but don’t know how to use Python or R. With more and more sources of information and a need to derive insights from expensive IoT deployments, no-code AI is another way to let more people participate in the next generation of technology innovation. (Stacey on IoT)
More trouble ahead for Nvidia’s plan to buy ARM: The U.S. Federal Trade Commission has “expressed concerns” over Nvidia’s $40 billion plan to buy chip architecture firm ARM. This isn’t a surprise, especially since regulatory agencies in the UK, China, and the EU have all expressed concerns about the deal. Because ARM designs an architecture that underpins almost every chip used in a smartphone and many used in the IoT, regulators and ARM’s customers are leery of a deal that would see Nvidia control that architecture. That’s especially true given that ARM has clear designs in the computing and server market, where Nvidia is also trying to sell its chips. Nvidia had expected the deal to close in early 2022, but in a conference call this summer, noted it was taking longer than expected. If it doesn’t close before the summer of 2022, ARM owner SoftBank gets to pocket a $1.25 billion breakup fee. (Protocol)
IoT platform Toit open sources its language: Is Toit the Kubernetes of the IoT? This August, I wrote about Toit, a company building software for the IoT that helps manage applications at scale. It’s one of several companies that are building an alternative to containers for the embedded devices used in IoT deployments. This week, the company said it would open source its Toit programming language, which lets developers build software that can run efficiently even on small chips. (Medium)
This is definitely a buyer beware situation: I say this all the time, but with connected devices, you aren’t buying hardware; you’re buying software. This is hard for people who are used to paying for a device and having it work mostly as intended in some static form for the rest of the life of the device. But in IoT, that’s not the case. If the company shuts down, your hardware might become a brick. If the company tweaks its business model, you might find yourself paying for features that used to be free. And in the case of NordicTrack X32i treadmill buyers, if you bought the device because it was easy to hack, that capability might one day disappear. NordicTrack customers apparently purchased the treadmill because it was easy to hack its 32-inch screen to play YouTube videos, Netflix, or whatever else they wanted in addition to NordicTrack classes. But a software update is breaking that feature and upsetting customers who paid $4,000 for a machine in part because they valued the ability to use its big screen to watch their own stuff. This is frustrating for the customers and not a great look for NordicTrack, and it just goes to show how far apart buyers and manufacturers are when it comes to understanding what a connected device actually is and how it should behave. (Ars Technica)
Another try for smart clothing: Over the years, Kevin and I have been excited by the potential and then disappointed with the limits of smart clothing. I’ve tried smart bras and smart socks, while Kevin has tried smart shirts. Each time, the potential to improve our form through tracking body movements or the ability to get detailed heart data has been cool, but not worth the high cost and frustration of charging the electronics that must be taken out before washing. But there’s another smart clothing brand on the market now, and diehards may want to give it a shot. The UK startup Prevayl is coming to the smart clothing market with an attractive shirt and an approach aimed at hardcore fitness users who are also into biofeedback. I’ll think of it as the Whoop contingent. (Engadget)
Another AI-based security startup for IoT has raised funding: Shield-IoT has raised $7.4 million in funding to help it build out sales and marketing. The four-year-old Israeli company differentiates itself from other startups using AI to detect anomalous behavior by saying it can do so at the scale of millions of devices. Doing anomaly detection can take time, especially as the number of networked devices grows (consider the difference between a lifeguard looking for someone drowning when there are four people in a pool vs. when there are 400). So instead of looking at individual behavior data, Shield-IoT builds a mathematical construct it calls a corset to shrink the data points in a way that doesn’t lose any crucial information and then runs its anomaly detection algorithms against that. I’m intrigued. (VentureBeat)