IoT news of the week for August 12, 2022

Arlo is making this subscription thing work: Last fall I wrote about Arlo’s financialsand wondered if the connected camera company could make it as a stand-alone business by forcing customers to pay for monthly subscriptions. It seems it’s doing well based on its results posted this week. During its second quarter, its annual recurring revenue hit $117 million, up 117% from the same period last year. It also has increased paid accounts year over year by 41%, saying that it now gets an average of $9.95 per month from subscribers, which translates to a lifetime value of $550 per customer. That covers the cost of the cloud subscriptions and then some. (Fool.com, Arlo)

Remember our concerns about ethical use of wearable sensor data? This article highlights all the ways the NBA is using newer forms of wearable sensors to track player health during practice (wearable sensors are not allowed during games per NBA rules). I know athletes are constantly seeking to optimize their performance while fighting against the limits of their health, but there’s something so cold about being able to accurately assess whether or not it’s worth the potential injury to score that point that it scares me. It feels close to hubris to think technology can actually determine these things and to try to engineer our way to the ideal physical outcome. After all, if both teams have access to the tech, it feels as if we would get to a place where we’re essentially watching a collection of performance optimization algorithms play a game against another set of algorithms. (FiveThirtyEight)

WiZ plans new lighting options for those on a budget: Signify, the company that owns both Philips Hue and the budget-conscious WiZ brand of connected light bulbs, said it will launch three new products, including a portable light, a smart button and bar, and floor lights for creating ambiance. WiZ lights use Wi-Fi instead of the Hue’s ZigBee+gateway option, and the two are not able to work together in the same app. (The Verge)

Why Amazon wants iRobot: Last Friday, Amazon said it would acquire iRobot, the company behind Roomba, for $1.7 billion. While many commenters focused on privacy, I’m more worried about how this deal could affect the development of a whole-home ecosystems comprised of many partners, as opposed to Amazon dictating all the moves. We’ll have to see if the FTC agrees; it’s certainly paying close attention to the impact such deals can have on the future of innovation. (Stacey on IoT)

Amazon lets more people pay with their palm: Amazon is rolling out palm payment readers at more Whole Foods stores, and I’m not sure how I feel about it. The Amazon One technology consists of a palm print reader at the entrance to stores. Users can upload their palm prints to the system and then hover them above the reader to get access to their Amazon account to pay. Using the system lets customers pay with their palm prints, and in certain stores using the Amazon Go cashierless checkout technology, could let customers go through and shop without needing to grab their wallet or their phone. I have used the Amazon Go stores and love the convenience of being able to walk out after picking up a few items, but I didn’t find the process of using the Amazon Go app and bar code scanner to enter the store so cumbersome that I’d need to switch to my palm print. I think a little friction is preferable to having Amazon store biometric information that’s irrevocably tied to me. (Ars Technica)

This is a nice security package focused on embedded systems: This links out to a list of different articles all focused on the challenges of IoT and embedded security at the moment. I found some to be helpful at indicating the attention the problem is receiving and the breadth of the challenge. For example, one article points out that shorter sales cycles for cybersecurity solutions in IoT environments are a sign of panic brought on by increased attacks. Another article lays out the potential new forms of attacks that embedded systems engineers have to account for when using higher-performance chips. Check it out if you’re new to the challenges embedded systems pose for securing devices. (EETimes)

Keeping models fresh and synthetic data are big trends in AI: I often say that the sensors and connectivity IoT systems comprise are the body, while AI is the soul. And when it comes to that soul, a survey from Appen, which provides products for those building and managing AI models, shows a few changes. One of the cooler changes for those of us excited about personalization is that 90% of those surveyed are retraining their models more than quarterly. This means those models should avoid drifting into inaccuracy and are taking advantage of newer data, hopefully improving the accuracy. Another trend is the increasing importance of synthetic data in training AI models. To build any model, a researcher needs data, but that data doesn’t have to be real. The interest in synthetic data to help round out weak points in existing models is essential and helps cut costs while also helping models recognize rare situations. (Appen)