IoT news of the week for May 17, 2019

Google smartly walks backs its Works with Nest transition: After stunning both its end users and device partners by announcing the Works with Nest program will shut down this August, Google this week did what it should have done in the first place: made clear that device integrations will not be broken during the transition, so long as customers keep their existing Nest account. This should allow approved integration partners to properly migrate over to Google’s new platform. However, new Google Home features won’t be available until customers migrate their Nest account to a Google account. (Google) — Kevin Tofel

There were a billion cellular IoT connections in 2018: Roughly 13% of the world’s cellular connections in 2018 were IoT-related, and that number will continue to grow, according to this report. However, revenue is forecast to grow much more slowly for IoT connections — and annual revenue per user (ARPU) is set to drop dramatically. Indeed, the more connections an operator has, the lower its ARPU goes. This isn’t surprising because connecting a sensor costs less per month than providing cellular connectivity for a cell phone. It’s also a trend in the chip world as well, specifically in the microcontroller space. Low-cost, low-margin connections are the wave of the future in the IoT, and companies have to adapt to this new model. (Berg Insights) — Stacey Higginbotham

Cambridge Consultants built a robot with better dexterity: Like babies, robots are pretty bad at picking things up. Historically, engineers programmed robots to handle a product of a specific size and shape. When faced with a new object, robots failed. But Cambridge Consultants has built a better robot, one that uses a sense of touch and slippage to adapt its pincers to the object it’s trying to grasp. While this sounds easy, it is actually a complex task that humans tend to take for granted. The key to development was giving the robot a feedback loop that let it “understand” how the object it wanted to grab was behaving as it grasped it. Commercializing this tech could lead to more generic robots that can take on more tasks in warehouses or even retail environments. (Cambridge Consultants) — Stacey Higginbotham

Translatotron marks an AI breakthrough in spoken language translation: Admittedly, I don’t take many calls on my Google Home device from people that speak a language other than mine. But after hearing the audio samples of Google’s Translation speech-to-speech model, I’m hoping to get more of such calls. The project comes by way of three Google teams: Brain, Speech, and Translate. It’s not yet as fast as traditional AI speech platforms, but what amazes me is how it attempts to replicate the original speaker’s actual sound. It’s not perfect, but it’s promising. (Google AI Blog) — Kevin Tofel

Amazon appears one step closer to robotic fulfillment: This story out of Reuters highlights Amazon’s latest move to use robots to pack items four to five times faster than humans, which even at a cost of $1 million each can save both money and time. While it’s easy to jump on the “robots are displacing human workers” bandwagon, let’s face it: That’s where Amazon is heading. And optimization and efficiency are what every company should be striving for. Moreover, these types of robots require both support and maintenance, often with a dedicated tech on site. Not every “product picker” can be one of those techs, but my hope is that as we automate physical business processes, companies such as Amazon look to train at least some of the displaced staff for a more enjoyable job as well as a higher salary. (Reuters) — Kevin Tofel

Japan is running out of phone numbers because of IoT: For several years, we’ve noted that billions of new IoT devices would be added to networks around the world. Now we’re seeing some of the negative effects, or at least Japan is. The Land of the Rising Sun is running out of 11-digit numbers due to the country’s requirement that even IoT devices have a unique phone number. As a result, Japan expects to release up to 10 billion new 14-digit numbers through 2021. Out of a special set of 11-digit numbers specifically for IoT devices, the country has already run through nearly half of the 80 million it launched just two short years ago (The Japan Times) — Kevin Tofel

Let’s use the IoT for marketing: This is a brilliant use of mobile GPS data, traffic data and quick-thinking algorithms. Burger King is offering people stuck in traffic the chance to order a Whopper from their phones (using voice) and Burger King will send a delivery person on a motorcycle to deliver said Whopper. Sure the connected “thing” in this example is a car or the phone, but it shows how ubiquitous connectivity and data crunching can change what’s possible. (The Spoon) — Stacey Higginbotham

A silly IoT bill has returned: A House Republican has reintroduced the SMART (State of Modern Application, Research, and Trends of) IoT Act that would force the Commerce Department to study the effects of IoT in the U.S.   The bill failed to pass the Senate last year and I wouldn’t mind if it failed again. I’d rather we focus on data protection regulations or better device security, as opposed to funding a survey. (MultiChannel News) — Stacey Higginbotham

Don’t let the machine take over just yet: As companies let machines take over more tasks, it can be tough to rectify the harms that can occur when the machines get things wrong. Everyone has probably had the fun of disputing a mistaken charge on a bill, but imagine the frustration when a machine decides you are a poor credit risk based on an opaque algorithm and you can’t get a loan. Or in this case, imagine when a car rental company has a “glitch” that reports your rental car stolen by mistake, and the police stop you to take it back. The resulting arrest records, fines and jail time seem like something that should have Hertz racing to fix this glitch or at least issuing some kind of notice to police forces that in the case of stolen rental cars, the police should ask more questions before making an arrest. (The Drive) — Stacey Higginbotham

Comcast’s MachineQ products are now available for all: This year at CES I talked to an executive at Comcast’s MachineQ connectivity business who told me that the company was going to focus less on providing LoRa-based connectivity to clients and more on delivering packaged IoT products to solve a particular need. It’s part of an ongoing trend to stop focusing on components of an IoT solution to instead deliver a solution. This is happening because many buyers of the technology found dealing with the technology too difficult, especially when all they wanted was a way to monitor water use on their golf course or some way of measuring the temperature of 100s of freezers across a chain of restaurants. And now, Comcast says that the MachineQ suite of products are now in use and available for folks who want to use connected things as opposed to building the infrastructure for those things. (Comcast) — Stacey Higginbotham

Are you interested in inference hardware at the edge? If this is you, click on over to see an excellent benchmarking effort done by my friend Alasdair Allan. He compared several of the ML accelerators boards such as the Google Coral board, the Nvidia Jetson Nano, the Intel Movidius, and even a Raspberry Pi using both TensorFlow and TensorFlow Lite. This is an excellent resource that digs deep. (Medium) — Stacey Higginbotham

Google Assistant on Sonos is mostly good but has a few compromises: This week, Sonos added Google Assistant to its Sonos One and Beam products, so I took the digital assistant for a spin on new hardware. Yes, it works mostly the same, but I learned a few things during my review. Namely, Alexa and Google Assistant can’t both be active on Sonos simultaneously, which might be a downer for some people. And a few Google Home-specific activities aren’t available — at least not yet — in the Sonos implementation. Check my review for the full story (StaceyOnIoT) — Kevin Tofel