Amazon now offers machine learning at the edge instead of in its cloud: Amazon Web Services detailed how customers can use AWS Outposts (which is a service that replicates AWS services from the cloud on a company’s own local gear) and Amazon’s SageMaker data analysis software to perform machine learning on a customer’s data locally, without sending it to the cloud. This is an essential element for clients who are worried about secret business processes getting exposed in the cloud and it’s also a pretty remarkable feat of computing to be able to handle actual learning at the edge. (AWS) — Stacey Higginbotham
SmartThings is finishing up its strategic shift toward the edge: SmartThings last year said it would stop letting developers build their own device handlers, and forced users away from an older version of its app while also stopping its commitment to making hardware. This week, it showed why it made those changes as it unveiled SmartThings Edge, a smart home automation system relying on others’ hubs, and a new SmartThings software that lets users build their own device handlers and share them among a community. (Stacey on IoT) — Stacey Higginbotham
Qualcomm unveils a 5G drone platform: Much like it did with the early-generation mobile phones, where it crammed the brains and radio of a smartphone on one single system on a chip (SoC), Qualcomm is cramming all of the 5G and AI onto an SoC for drones. Qualcomm just announced its Qualcomm Flight RB5 5G Platform, which uses the Qualcomm QRB5165 processor. The result of all of this integration should be a smaller silicon footprint that uses less power than non-integrated chips. That’s good for drones, and essential for Qualcomm, which needs to find markets to replace the cell phone since its biggest customers are now making their own application processors and working on their own radios. (Qualcomm) — Stacey Higginbotham
We’re giving AI waaaay too much credit here: I loved this article because it brought up one of my favorite Silicon Valley tropes. The fear that the mighty algorithm, be it in search, social media, or in smart cities, will give us exactly what we think we want and never surprise us. This article is all about how smart cities could have the potential to become boring because the systems dictating where we go and what we see will become so attuned to our preferences that we will never leave our comfort zone. I suppose a brand-new from scratch city might be able to exclude the messiness of other humans and older generations of city life, but I have faith in a mashed-up version of the future where art, other people, and other perspectives still have a chance to intrude and keep us from becoming bored automatons. I think a more realistic fear is that a smart city has so much information about us that we fear becoming anything more interesting than bored automatons. (PCMag) — Stacey Higginbotham
Rapid Robotics grabs $36.7M in funding: Only two years after its founding, startup Rapid Robotics has raised its third round of funding. That sounds fast to me, but the “rapid” part of its name really has to do with how quickly it can deploy robots. Its robotic arm can be up and running in hours thanks to pre-programming and AI models. And according to the company, it can accomplish 80% of tasks typically required on a factory floor. (Silicon Angle) — Kevin C. Tofel
So long, Intel intelligent cameras; it’s been Real…Sense: Six years after debuting its RealSense camera hardware, Intel is saying goodbye to it. The company this week announced it would wind down RealSense operations and reallocate its resources. I can’t say I’m surprised, as I’m betting more than half of you are saying “Real what?” The camera hardware never really took off, with reports of just 10 customers buying it in low volumes. Indeed, I saw very few end-user products that took advantage of the RealSense camera’s combination of AI, depth sensing, and 3D imaging. (CRN) — Kevin C. Tofel
Yes, attractive colored lights can be Thread routers: Nanoleaf continues to lead the way when it comes to integrating Thread technology into its products. The smart lighting company this week announced that its existing controllers for the Nanoleaf Shapes and Elements lights can act as Thread Border Routers. The functionality comes by way of a software update for the controllers. Since these are always plugged into an outlet, they’re the perfect candidate for Border Router functionality; they can act as gateways or bridges between data on Thread networks and the internet. The updated controllers join Apple’s HomePod mini and eero routers as Thread Border Routers currently available. (MacRumors) — Kevin C. Tofel
How to use machine learning at the edge on a Pi: I love stories like this that share a step-by-step process for building your own smart devices. This one combines a low-cost Raspberry Pi, Edge Impulse services, and a webcam. Along with a few light Python scripts, you can combine these three things into a simple missing lug nut detector when the camera sees the wheel of a vehicle. It may not be the most practical application, but that’s not the point. Once you learn the basics here, you can start to tinker and tweak things to design some other smart system. (Raspberry Pi Blog) — Kevin C. Tofel
It’s not easy being green: I worry a lot about the environmental impacts of the IoT from obsolete consumer gadgets to discarded industrial sensors, so I was excited to read my former colleague’s series on how consumer companies are trying to think about designing sustainable gadgets. I liked the updates on how big tech brands were developing sustainable materials and packaging and was stressed to see that it’s so much more difficult to take it much beyond that level without new materials and technology. IT probably means I should buy fewer gadgets until they work this out. (Protocol) — Stacey Higginbotham
Leave a Reply