|
Did you receive this newsletter as a forward? Subscribe here
|
|
| Where does generative AI fit in the IoT? |
| |
| By Stacey Higginbotham |
Major publications are really worried about Alexa, Siri, and Google's digital assistant these days. That's because, so far, these products aren't using generative AI, a set of machine learning models that are trained to guesstimate the next set of words or the right image to create based on a prompt.
DALL-E, Stable Diffusion, and Midjourney are popular on the image generation side, and came out back in early 2021. But the latest hype is around large language models, specifically ChatGPT created by OpenAI. When someone puts in a prompt and a style suggestion, the results can make easy reading and give actual information or the illusion of information. |
|
| — I asked Stable Diffusion for an image of an AI talking to the internet of things. It's clear that stock art heavily influenced this prompt. |
|
|
All of these tools are exciting and will have a significant impact on how work gets done, how people create content, and how we build businesses. And yes, there are concerns over accuracy, the ability to use generative AI for easy propaganda, scams, and other malicious acts that we still must grapple with rather than simply blindly accepting that generative AI is the messiah technology that will Change The World.
That said, right now, we're at the stage where technologists are creating blind belief in generative AI, including by putting billions of investment dollars into companies experimenting with use cases and new models. We're also at the stage where members of the media spend hours trying to trick these models into behaving badly or try to prove that the AI is sentient and may have hostile feelings toward us.
I'm not really here to talk about that. I want to focus on areas where generative AI will have a significant impact on the way the Internet of Things gets deployed and used. For example, where can we use it to improve user experiences? What types of jobs might it aid or take over? And words and images aside, what other generative AI models might help with the IoT?
I'll start in the smart home, because that is where my heart is. Rather than confuse Amazon Alexa with what is essentially a super advanced chatbot, and predict that Alexa will "lose the AI race" or have its thunder stolen, it's likely that it and other digital assistants will continue to use natural language processing (NLP) for "understanding," then taking action on, various task-based requests like "Turn on the lights," or "Good morning" to start a wake-up routine, while also adding in a GPT-style chatbot to handle requests that require more in-depth communication.
A good digital assistant isn't going to have one or two models, but be comprised of whatever version of models provides the most utility for the user. There are also economic concerns. Calling out to a chatbot might incur charges that require a different business model, and not everyone will want to pay a subscription to get Siri to tell them a story or share with them the best recipes based on the food in their fridge.
Plus, sooner rather than later we will see chatbot-style generative AI models in use in the home. I recently had Paulus Schoutsen, the founder of Home Assistant, demo how to use a HomePod to access a GPT-style chatbot to tell his kid a story. And I think being able to combine the NLP already in use inside a smart speaker to translate my spoken words into written prompts for a GPT-style model could make for an easy way for me to describe a routine for Siri and then write the code or connect the integrations needed to implement it.
Indeed, I think the utility of using both the NLP that's already part of digital assistants in combination with a generative AI model is clear to SoundHound, which is introducing a platform that combines a voice assistant with a generative AI. So ChatGPT won't kill Alexa, but it will probably end up becoming part of Alexa, with Alexa as the interface and ChatGPT just one of many services it provides.
Other areas in the smart home where ChatGPT or generative AI models will have an impact include children's toys, fitness services (have the model deliver a custom workout), recipes, or suggestions of things to do. That's because generative AI is really just another reason to add connectivity and sensing to everyday objects, either to provide personalized training data or act as a conduit to such services.
On the enterprise side, there's the obvious utility of using generative AI to help business people implement digital solutions without coding. One example is how Software AG has combined its webMethods cloud-to-cloud integration platform with a generative AI model to help employees figure out how to link data and various digital services. Eventually, as we connect more things in buildings, manufacturing lines, commercial kitchens, etc., using plain written language to tell our connected devices how to work with our connected business software will help managers become more efficient and capable.
And in industrial environments, the promise of ChatGPT comes with compelling use cases and caveats. Several people have championed using generative AI for things like predictive maintenance. Generative AI models work by training on massive quantities of data and then generating the most likely next element. So in large language models, the generative AI model is training on huge swaths of text and generating what the model thinks is the most likely next word or phrase.
Presumably, with enough machine data, a model could decide what's supposed to be next and send an alert if the expected result isn't right. But frankly this feels like overkill, since traditional anomaly detection works fine for predictive maintenance and is much less cost-intensive. Where generative AI might get interesting is by taking process data and suggesting alternative workflows, or by using written language to describe workflows and having an AI code it for someone.
But there are caveats. These models are only as good as their training data and in some cases can generate wrong answers, but can be written so well that it's hard to determine if they're wrong.
"If you ask the technology to provide possible answers to technical questions, our experience is that, without proper context setting and filters, 80% of the answers are not accurate — potentially even harmful," said Erik Udstuen, CEO of TwinThread via email. "But with proper context setting and filters/guardrails you can get very high accuracy."
TwinThread uses a variety of different AI models to provide customers with services based on digital twins of their physical operations, including using generative AI to provide vast quantities of information to frontline workers and managers when they need it. Think of asking the AI why a specific batch of chemicals is too acidic or some other question related to a specialized process.
Udstuen believes the challenges around accuracy are a short-term issue that vendors will figure out, and that a bigger challenge with generative AI is related to intellectual property and the perceptions around how generative models are trained. "The last thing that anyone wants is to provide proprietary context or information that comes to be part of the public domain," he told me.
He added that some customers are so worried about IP leaking they ask to turn any sort of ability to ask a chatbot about their specific processes off. While it's not the case that their questions or proprietary data is shared, it's such a big perceived risk for customers they don't want to risk it.
Given the IP fights around generative AI, this last concern "feels" like it would be a problem. But in reality it is relatively simple to set limits on where training data actually comes from — or even if a model built on proprietary data gets deployed outside of the intended plant or business.
Time and education about how generative AI models are created and how they work will address some of the IP concerns. And since we're only a few months into this hype cycle, I have faith that we're going to see generative AI become as important as computer vision and NLP, and as accepted.
We're also going to see some interesting new use cases for the IoT, so feel free to share what y'all are doing and thinking. |
|
|
|
|
| How Telecoms Reduce Network Downtime SPONSORED |
| |
|
Telecommunication organizations need to ensure they have the necessary resources and technology to maintain service uptime SLAs.
Increased regulations and emerging technologies forced telecommunications companies to evolve quickly in recent years. These organizations’ engineers and site reliability engineering (SRE) teams must use technology to improve performance, reliability and service uptime. Learn how WideOpenWest uses a time series platform to monitor its entire service delivery network.
Read full story |
|
|
|
|
| It's time for a smart home spring clean, y'all! |
| |
Last week, I had three different batteries on my desk and a growing list of tasks I needed to do to keep my devices online and running smoothly. After swapping out batteries in a door lock, a door knob, and in the sensor of the MyQ garage door opener, I got another low battery notice from a Hue motion sensor in the storage room.
And I still need to clean the camera in my oven before it stops recognizing the food I put in and recommending the perfect cook time. I'm also worried that if I don't clean the sensors on my Roomba it might take a dive off the stairs or simply stop cleaning as effectively. Plus, I have a number of devices that have fallen offline and need me to reconnect them, and a pair of Nest Audio speakers that just need to get recycled because they continuously shout, "I can't help with that!" if anyone asks any other Google speaker for anything.
It is, in other words, spring cleaning time. |
|
| — One of my tasks is cleaning the camera on the June oven, to ensure it can still recognize my food. Image courtesy of S. Higginbotham. |
|
|
This is not just a litany of annoying small projects related to my many connected gadgets that I've completed or need to complete soon. It is a very detailed list of things that I try to cluster around this time of year for the annual smart home spring cleaning that I do. In some houses, people might wash their walls or clean their garage. In my home, it's time to gather a smartphone, a cleaning rag (for those dirty cameras and sensors), and a truly prodigious supply of batteries, and refresh, reset, or reject each of the devices in my home.
This year, I had thought that I'd be doing a massive reset of my smart home hubs because I'd be upgrading everything to Matter, but that's not happening just yet. So instead I'm going through and making sure my naming conventions are all in sync across the four home hubs I have running (Amazon, Google, SmartThings, and an undisclosed review unit). I make sure that every room in the house has the same name and that each device in each room has the same name (which is not the name of the room unless all of the devices in that room get turned on all the time). We talk about how to name devices in this post here, and why you shouldn't get cute with your smart home names here.
I also check that my devices are all connected and reporting in. And that my automations still work for the life we're leading today. (If you have a lot of different options on your devices, you can check this by seeing which automations have run recently). I also make sure the automations power the right devices. I swap out devices often, so a goodnight routine that turns off a door lock that's no longer installed isn't helpful.
I also open each smart home device app on my phone to check for messages, updates, and that I can still use it the way I want. For example, at some point in time the makers of my oven decided that I needed to manually activate the ability to remotely control the oven on my app by physically hitting buttons on the oven to enable the feature. This is probably a smart decision, but it is nice to know this before I try to preheat my oven away from home.
To be sure, it's a lot of fiddly work. And while things like Matter will make this process easier in principle by carrying devices and their names across my different systems, and presumably making it easier to get devices back on the network if they fall off, my automations will still have to be checked and tweaked, and batteries will still need changing, and sensors will still need cleaning. And yes, I'll still have to open 50 different apps.
If I wanted to take my spring cleaning a step further I'd also check the security stats of my devices, making sure the software was up to date (mine are all set to automatically update) and look at the data traffic patterns on my Firewalla to see if anything is behaving oddly. (Although I am fairly certain I'd get a notification if something was odd.)
Spring cleaning is also a great time to check passwords — and for anyone not using a password manager, taking the time to set one up. I think password services are essential for smart home users because we're constantly linking services together through new hubs or new cloud integrations. Just the other day when setting up a new hub I had to enter almost a dozen passwords before I was able to deem the setup good enough and stop adding devices. Multi-factor authentication on devices where that makes sense is another must-have. Cameras, devices that control systems that could cause big damage such as HVAC and water, and appliances such as ovens or dishwashers are all good candidates.
I also use this time to look at any of my device subscriptions to figure out what I should or shouldn't be paying for going forward.
The process isn't fun, and some people are so organized they probably just fix problems the moment they pop up on their smart home network. But for me, especially when it comes to smaller devices I don't rely on every day, it feels good to designate a time of year (and sometimes two) when I take stock and deal with anything that's annoying but not so annoying that I fixed it the moment I first saw it.
If you have your own smart home spring cleaning tasks, I'd love to hear about them. |
|
|
|
|
| Bluetooth Technical Training Series SPONSORED |
| |
|
Join Silicon Labs' expert engineers for a series of one-hour virtual technical trainings covering the most important aspects of leveraging Bluetooth for IoT devices.
Our Bluetooth Tech Talks series covers a wide range of topics like machine learning, what's new with Bluetooth Mesh and HADM, and provides hands-on development training for creating reliable, secure connected devices.
Register now to tune in, ask questions, and accelerate your device development today. |
|
|
|
|
|
|
| Why did the average number of smart home devices per home drop? |
| |
|
| — The average smart home has 7.1 smart home devices, according to Parks Associates, which released new data this week. The research firm does clarify that when we look at all U.S. homes, only 28% have three or more connected smart home devices, which presumably qualifies them as smart homes. Folks like me, with my 60+ connected devices, are counteracting a crop of new buyers who have one or two smart home devices, putting the average number of device in smart homes on the decline. Parks also notes that the average number of devices has dropped a bit since the end of 2021 due to inflation. I wonder if people waiting until Matter products hit the market to buy devices also slowed adoption. |
|
|
|
|
|
|
|
| News of the Week |
| |
Arm may be trying to change its pricing model: The Financial Times reports that Arm is asking customers to change the way they pay for the chip design licenses they buy from Arm. Arm offers two types of licenses; one is a simple fee per chip to use an existing design, and the other is a larger fee to take the initial design and tweak it for special purpose. The FT is reporting that Arm is asking its largest customers to change the per-chip fee to a new fee structure based on the selling price of the device the chip ends up in. (The article does note that Apple is not part of these particular price discussions, despite having an Arm architecture license to build chips used in MacBooks and iPhones.) So a smart phone or computer using the ARM architecture might cost more in fees than a Wi-Fi router. The shift in pricing comes after a failed deal to sell Arm to Nvidia and ahead of a planned IPO. My hunch is that Arm, which has seen its chips become the bedrock of so much computing, thinks its time to extract more value. It has tried with cloud-based services, which didn't pan out, and now is seeking a shift to generate higher revenue. Arm, like all chip companies, also spends more on software and security, and likely wants to recoup some of that. It also has to provide a return to Softbank, which purchased Arm for $32 billion in 2016. The pricing plans are likely to push some of its customers to alternative architectures such as RISC-V or specially designed FPGAs, but those moves will take a few years. (FT)
IFTTT is adding generative AI to its services: If you are still on Twitter, you may have seen that I am performing a small experiment where an AI generates tweets each time I publish a blog post. I'm using a new service from IFTTT (it costs $6 a month for the class of services that include the AI products) that has an generative AI based on GPT-3 create short-form social posts, long-form content like outlines or blog posts, or summaries of existing content like meeting notes or blogs. As you can see from my experimental tweets, it's pretty good at the summary, but it fails to include a link back to my story, which makes it useless as a means of driving traffic back to my site. I don't plan on using it to write my blog posts, but it might come in handy for folks who want to get a summary of my incredibly long newsletters. (IFTTT)
3 startups offering connectivity while sipping power: I love writing about radio options that can work on harvested energy or batteries (without requiring constant battery changes). This article highlights three different startups that can provide connectivity with low-power budgets. Go read about Spark Microsystems, Ixana, and Nanopower Semiconductor. (EETimes)
Millimeter wave sensors are the next big sensor for elder care: I read a lot about cameras or robots designed for senior living homes or as tools for monitoring the elderly, but my hunch is that the humble millimeter wave sensor will win out over these higher-tech (and more expensive) solutions. The latest proof point is research from the University of Waterloo showing a millimeter wave (mmWave) sensor system that uses disruptions in radio waves to detect motion, and AI to translate those disruptions into healthy movement, presence, and fall detection. In a few years, every senior living facility will likely have a dedicated mmWave sensor on the wall or ceiling, or perhaps include them as part of wireless access points. (University of Waterloo)
And mmWave sensing isn't just for older folks: This project, called RoomSense IQ, uses a variety of sensors to "understand" what's going on in a room, but the primary detection mechanism is a mmWave sensor for motion and presence detection. The project also includes a more traditional PIR sensor that is used as an additional data source to prevent false positives. There is also a temperature, humidity, and light sensor. So while I'm waiting for mmWave sensors such as the really cool one we saw at CES from Aqara, I could build this one to try today. (HackaDay)
Aqara launches its first HomeKit Secure Video doorbell: Fans of the HomeKit Secure Video have had a limited selection when it comes to video doorbell options. The Logitech Circle View Doorbell and the Wemo Smart Video Doorbell were basically it. But now you can spend $199.99 on the Aqara G4 video doorbell and get support for HomeKit Secure Video. The doorbell can be installed with our without wires and requires six AA batteries. It provides 1080p video and a 7-day record of events without a subscription. (Aqara)
Renesas acquires NFC chip company: Chip firm Renesas has agreed to acquire Panthronics AG, a fabless semiconductor company specializing in Near Field Communication products, in an all-cash transaction. The amount of the deal was undisclosed. Panthronics makes NFC chips used in asset tracking, payments, wireless charging and automotive applications. I've been seeing some really novel uses of NFC in IoT products for access control that use NFC for communications and for wireless charging. We're also seeing NFC used for digital wallets and payments across a wide variety of wearables and phones. The two companies have been partners on products in the past, and the deal is expected to closed by the end of 2023. (Renesas)
Nanoleaf's Matter lights are here but other Matter gear isn't: Nanoleaf has released its Matter-compatible line of light bulbs and light strips. These are some of the first Matter devices to launch that aren't smart plugs, so we're pretty excited. Also, the Nanoleaf Essentials bulb that provides colors and white light costs $19.99, which is a great price for a connected color bulb. For now, they will work only if you have a Google or SmartThings Matter controller. (Stacey on IoT)
Want to sponsor this newsletter and the IoT Podcast? Our 2023 media kit is now available. Request a media kit.
|
|
|
|
|
|
|
|
|
|
