Apple’s long-rumored Mac ARM chip transition could happen as early as next year, according to a new report from Bloomberg. The report says that Apple is currently working on three Mac processors based on the design of the A14 system-on-a-chip that will power the next-generation iPhone. The first of the Mac versions will greatly exceed the speed of the iPhone and iPad processors, according to the report’s sources.
Apple and CMU researchers demo a low friction learn-by-listening system for smarter home devices
A team of researchers from Apple and Carnegie Mellon University’s Human-Computer Interaction Institute have presented a system for embedded AIs to learn by listening to noises in their environment without the need for up-front training data or without placing a huge burden on the user to supervise the learning process. The overarching goal is for smart devices to more easily build up contextual/situational awareness to increase their utility.
The system, which they’ve called Listen Learner, relies on acoustic activity recognition to enable a smart device, such as a microphone-equipped speaker, to interpret events taking place in its environment via a process of self-supervised learning with manual labelling done by one-shot user interactions — such as by the speaker asking a person ‘what was that sound?’, after it’s heard the noise enough time to classify in into a cluster.
A general pre-trained model can also be looped in to enable the system to make an initial guess on what an acoustic cluster might signify. So the user interaction could be less open-ended, with the system able to pose a question such as ‘was that a faucet?’ — requiring only a yes/no response from the human in the room.
Refinement questions could also be deployed to help the system figure out what the researchers dub “edge cases”, i.e. where sounds have been closely clustered yet might still signify a distinct event — say a door being closed vs a cupboard being closed. Over time, the system might be able to make an educated either/or guess and then present that to the user to confirm.
They’ve put together the below video demoing the concept in a kitchen environment.
In their paper presenting the research they point out that while smart devices are becoming more prevalent in homes and offices they tend to lack “contextual sensing capabilities” — with only “minimal understanding of what is happening around them”, which in turn limits “their potential to enable truly assistive computational experiences”.
And while acoustic activity recognition is not itself new, the researchers wanted to see if they could improve on existing deployments which either require a lot of manual user training to yield high accuracy; or use pre-trained general classifiers to work ‘out of the box’ but — since they lack data for a user’s specific environment — are prone to low accuracy.
Listen Learner is thus intended as a middle ground to increase utility (accuracy) without placing a high burden on the human to structure the data. The end-to-end system automatically generates acoustic event classifiers over time, with the team building a proof-of-concept prototype device to act like a smart speaker and pipe up to ask for human input.
“The algorithm learns an ensemble model by iteratively clustering unknown samples, and then training classifiers on the resulting cluster assignments,” they explain in the paper. “This allows for a ‘one-shot’ interaction with the user to label portions of the ensemble model when they are activated.”
Audio events are segmented using an adaptive threshold that triggers when the microphone input level is 1.5 standard deviations higher than the mean of the past minute.
“We employ hysteresis techniques (i.e., for debouncing) to further smooth our thresholding scheme,” they add, further noting that: “While many environments have persistent and characteristic background sounds (e.g., HVAC), we ignore them (along with silence) for computational efficiency. Note that incoming samples were discarded if they were too similar to ambient noise, but silence within a segmented window is not removed.”
The CNN (convolutional neural network) audio model they’re using was initially trained on the YouTube-8M dataset — augmented with a library of professional sound effects, per the paper.
“The choice of using deep neural network embeddings, which can be seen as learned low-dimensional representations of input data, is consistent with the manifold assumption (i.e., that high-dimensional data roughly lie on a low-dimensional manifold). By performing clustering and classification on this low-dimensional learned representation, our system is able to more easily discover and recognize novel sound classes,” they add.
The team used unsupervised clustering methods to infer the location of class boundaries from the low-dimensional learned representations — using a hierarchical agglomerative clustering (HAC) algorithm known as Ward’s method.
Their system evaluates “all possible groupings of data to find the best representation of classes”, given candidate clusters may overlap with one another.
“While our clustering algorithm separates data into clusters by minimizing the total within-cluster variance, we also seek to evaluate clusters based on their classifiability. Following the clustering stage, we use a unsupervised one-class support vector machine (SVM) algorithm that learns decision boundaries for novelty detection. For each candidate cluster, a one-class SVM is trained on a cluster’s data points, and its F1 score is computed with all samples in the data pool,” they add.
“Traditional clustering algorithms seek to describe input data by providing a cluster assignment, but this alone cannot be used to discriminate unseen samples. Thus, to facilitate our system’s inference capability, we construct an ensemble model using the one-class SVMs generated from the previous step. We adopt an iterative procedure for building our ensemble model by selecting the first classifier with an F1 score exceeding the threshold, 𝜃&'( and adding it to the ensemble. When a classifier is added, we run it on the data pool and mark samples that are recognized. We then restart the cluster-classify loop until either 1) all samples in the pool are marked or 2) a loop does not produce any more classifiers.”
Privacy preservation?
The paper touches on privacy concerns that arise from such a listening system — given how often the microphone would be switched on and processing environmental data, and because they note it may not always be possible to carry out all processing locally on the device.
“While our acoustic approach to activity recognition affords benefits such as improved classification accuracy and incremental learning capabilities, the capture and transmission of audio data, especially spoken content, should raise privacy concerns,” they write. “In an ideal implementation, all data would be retained on the sensing device (though significant compute would be required for local training). Alternatively, compute could occur in the cloud with user-anonymized labels of model classes stored locally.”
You can read the full paper here.
Samsung’s Galaxy Watch blood pressure monitoring app approved by South Korean regulators
Samsung Electronics announced today that its blood pressure monitoring app for Galaxy Watches has been approved by South Korean regulators. Called the Samsung Health Monitor, the app will be available for the Galaxy Watch Active2 during the third quarter, at least in South Korea, and added to upcoming Galaxy Watch devices.
TechCrunch has contacted Samsung for more information on when the app, which uses the Galaxy Watch Active2’s advanced sensor technology, will be available in other markets.
It was cleared by South Korea’s Ministry of Food and Drug Safety for use as an over-the-counter, cuff-less blood pressure monitoring app. The app first has to be calibrated with a traditional blood pressure cuff, then it monitors blood pressure through pulse wave analysis. Users need to recalibrate the app at least once every four weeks.
According to a recent report by IDC, in the last quarter of 2019, Samsung wearables ranked third in terms of shipments, behind Apple and Xiaomi, with volume driven by its Galaxy Active watches. Samsung has sought to differentiate its smartwatches with a focus on health and fitness monitoring, including sleep trackers.
VanMoof introduces new S3 and X3 electric bikes
VanMoof is releasing a new generation of its electric bike. In many ways, the VanMoof S3 and its smaller version the VanMoof X3 are refined versions of the VanMoof Electrified S2 and X2. It features an updated motor, hydraulic brakes and a familiar design.
If you’re not familiar with VanMoof bikes, the company has been building electric bikes with some smart features, such as an anti-theft system. There’s an integrated motion detector combined with an alarm, a GPS chip and cellular connectivity. If you declare your bike as stolen, the GPS and cellular chips go live and you can track your bike in the VanMoof app.
The company wants to control as much of the experience as possible, which means that it designs the bike in house, sells it on its website and in its own stores. 80% of orders happen on the website and VanMoof now has nine stores around the world. The company has sold 120,000 bikes over the years.
The S3 and X3 still feature the iconic triangular-shaped futuristic-looking frame. The electric motor has been updated — it is more powerful, more responsive, quieter and smaller. You’re not going to constantly switch from one gear to another as there’s an electronic gear shifting system — it has been updated from two gears to four gears. All you have to do is jump on the bike and start pedaling.
A big improvement compared to the previous generation is that the S3 and X3 now feature hydraulic brakes instead of mechanical disc brakes. And you’ll find the good old boost button on the handlebar to get an extra burst of acceleration when you need it.
When it comes to design, the saddle has been redesigned, the coating on the bike is now matte and you’ll see a lot of changes across the board. The only difference between the S3 and X3 is that the S3 is designed for taller people while the X3 is designed for smaller people. Unfortunately, it looks like the battery is still not removable.
The company is trying to control the supply chain as much as possible. It works with a small set of suppliers to manufacture custom components and then tries to cut out as many middleperson as possible to bring costs down. The VanMoof S3 and X3 cost $1,998/€1,998 but the company could raise the introductory price in the future due to pressure on the supply chain.
Here’s a video of the previous generation VanMoof Electrified X2 we shot a couple of months ago:
3D-printed glasses startup Fitz is making custom protective eyewear for healthcare workers
A lot of startups have answered the call for more personal protective equipment (PPE) and other essentials to support healthcare workers in their efforts to curb the spread and impact of COVID-19. One of those is direct-to-consumer 3D-printed eyewear brand Fitz, which is employing its custom-fit glasses technology to build protective, prescription specs for front-line healthcare workers in need of the best protection they can get.
The $99 Mendel Air Sensor uses data to help you grow better veggies (or weed)
The Mendel Air Sensor app is the first app I open every morning. Before Reddit, before Gmail, before NYT. I roll over, grab my phone and check my plants. I don’t know if there’s a higher honor I can bestow on an app.
The Mendel Air Sensor is a game-changer for indoor growers. It offers a sophisticated suite of sensors that collects critical information about growing conditions. With a price of $99, there’s very little else on the market that offers the same sort of data collection at an affordable price.
The company behind the Mendel Air Sensor started by building similar sensors for at-home aquariums. This group knows data collection and teamed up with an experienced manufacturer to develop and ship the Mendel Air Sensor.
I know very little about growing plants indoors. I’ve watched some YouTube videos, read a lot of blog posts and asked friends for advice. And yet I have a small growing operation in my basement: tomatoes, romaine lettuce, carrots and, you know, other leafy greens.
Several weeks in, I’m starting to appreciate the data behind growing plants. There’s a lot to consider, from the temperature to types and amount of light, to humidity and how the plants react to humidity through a calculation to determine the vapor pressure deficit (VPD).
I have a Mendel Air Sensor hanging in one grow tent (pictured at the top), and it’s my new obsession. The small green device collects four data points every 15 minutes and displays the information through a web app or smartphone app. This is allowing me to fine-tune the controlled environment through exhaust fans, light placement and humidifier levels.
As I’ve found, it’s critical to watch this data throughout the day. I’ve yet to stabilize the environment to a point where I set it and forget it. About twice a day, because of the Mendel Air Sensor, I make slight changes to the growing tent, which results in dramatic changes to the environment. Without access to this data, I wouldn’t know something is off until the plant shows warning signs — and as I understand it, that’s when it’s too late.
At $99, it’s a good value, and there are only a few competitors in the space. Most are double or triple the price, though their charting products seem more mature.
CEO Nate Levine tells TechCrunch Mendel started as a 50/50 partnership with another bootstrapped company, RapidLED out of the Bay Area. This company has sold lights for indoor growers for the last few years and already has an established base of customers in this field. But Levine didn’t start to build a product for monitoring plants; instead, he created, FishBit, a product for monitoring aquariums.
The parallels between the two markets helped Levine’s team jump into the indoor gardening space. As Levine told TechCrunch, the consumer demands are similar, and like with aquariums, indoor growers are increasingly looking for ways to increase capabilities. Instead of keeping fish alive, though, they’re trying to get more tomatoes. Or weed.
Levine said that unlike with aquariums, indoor growers can be less stingy with their cash, though, right now, with cannabis, margins are slim. There isn’t a gold rush, he said, but noted that the cannabis market, in particular, is at the right spot for companies to launch new products.
The company is marketing the same product to home growers, and commercial growers thought this could be a challenge with the current web app. It lacks robust features found on other products. For a small grower like me, it’s okay, but I expect commercial customers expect better logging, more detailed analysis and a variable monitoring cycle instead of just every 15 minutes.
To make it available for international users, the company needs to swap out the USB power supply.
Don’t call this is a pivot. Or at least Levine doesn’t call it a pivot. As he told TechCrunch, if he goes back to the original pitch deck, the company is still driving at the same goal for FishBit, and everything the team learns on Mendel is implemented in FishBit, too. The goal is to build an entire product line of smart hardware and software for the indoor grower.
RapidLED approached Levine and the team at an aquarium conference and offered to build the hardware if Levine could make the software. My plants are happy that the two companies forged the partnership.
As for my plants, I’ve learned a few things because of the Mendel Air Sensor. First, my grow lights put out much more heat than I expected, and I need to dump the cheap set and get a name brand unit. Second, the humidity was much lower than I had expected, so I added a humidifier. Finally, monitoring the VPD is much easier than it seems if the calculations are automated.
Growing plants is hard, but it’s easier with the data from the Mendel Air Sensor.
Apple said to be working on modular, high-end, noise-cancelling over-ear headphones
Apple is said to be developing its own competitors to popular over-ear noise-cancelling headphones like those made by Bose and Sony, Bloomberg reports, but with similar technology on board to that used in the AirPod and AirPod Pro lines. These headphones would also include a design with interchangeable parts that would allow some modification with customizable accessories for specific uses like workouts and long-term wear, for instance.
The prototype designs of the new headphones, which are set to potentially be released some time later this year (though timing is clearly up in the air as a result of the ongoing COVID-19 crisis, and Apple’s general tendency to move things around depending on other factors), are said to feature a “retro look” by Bloomberg, and include oval ear cups which connect directly to thin arms that extend to the headband. The swappable parts include the ear pads and headband cushion, both of which are said to attach to the headphone frame using magnetic connectors.
These will support Siri on board, along with active noise cancellation and touch controls, but most importantly for iOS and Mac users, they’ll also feature the simple connection across multiple devices that are featured on AirPods and some of Apple’s Beats line of headphones.
Apple has already released Beats over- and on-ear headphone models with AirPod-like features, including cross-connectivity, and that feature onboard noise cancellation. The Bloomberg report doesn’t seem to indicate these new models would be Beats-branded, however, and their customization features would also be new in terms of Apple’s available existing options.
Bloomberg also previously reported that Apple was working on a smaller HomePod speaker as part of its forthcoming product lineup, and a new FCC filing made public this week could indicate the impending release of a success to its PowerBeats Pro fully wireless in-ear sport headphones.
GoPro lays off 200 employees representing 20% of the company
Action camera manufacturer GoPro has announced some massive organizational changes at the company. In particular, the company is laying off more than 200 employees — which represents a 20% staff reduction.
GoPro plans “office space reductions in five geographies” as well as a reduction in operating expenses. The company expects a “0 million reduction in non-GAAP operating expenses in 2020 and plans to further reduce operating expenses into 2021 to 0 million.” The 2021 reduction in operating expenses will come from “non-headcount related operating expenses.”
Behind the scene, GoPro is making some radical changes to its business model. The company is still selling cameras, accessories and subscriptions. But it is switching to a direct-to-consumer model with GoPro.com acting as the main storefront.
The company will stop selling its devices in many retail stores. GoPro will still work with select retailers in some regions as they still generate a lot of sales.
But selling directly on GoPro.com represents the future of the company, which should improve margins. They don’t have to give a cut to retailers when GoPro is acting as the retailer. In 2019, GoPro.com represented a bit more than 20% of European revenue and a bit less than 20% of revenue in the U.S.
“GoPro’s global distribution network has been negatively impacted by the COVID-19 pandemic, driving us to transition into a more efficient and profitable direct-to-consumer-centric business over the course of this year,” founder and CEO Nicholas Woodman wrote. “We are crushed that this forces us to let go of many talented members of our team, and we are forever grateful for their contributions.”
In addition to today’s layoffs, sales and marketing expenditure will be cut down in 2020 and beyond. Woodman himself won’t take a salary for the remainder of the year. The company’s board won’t receive any cash compensation either.
GoPro has withdrawn earnings guidance for Q1 and 2020. The company now expects revenue of $119 million with a non-GAAP EPS loss of $0.30 to $0.40 per share. In pre-market trading, GoPro shares are currently trading at $2.58, 3% below yesterday’s closing price.
Correction: An earlier version of this story mistakenly stated that GoPro will only keep offices in five geographies and that the reduction in operating expenses in 2021 could involve further layoffs. Instead, the company is reducing its real estate portfolio in five geographies and the 2021 reduction in operating expenses will come from non-headcount related operating expenses.
FDA authorizes production of a new ventilator that costs up to 25x less than existing devices
The U.S. Food and Drug Administration (FDA) has authorized the manufacture of the Coventor ventilator, a new hardware design first developed by the University of Minnesota. The project sought to create a ventilator that could provide the same level of life-saving care as existing ventilator models, but with a much lower cost to help ramp production quickly and make them affordable to the health institutions that need them.
Electrical worker safety startup launches a COVID-19 workplace distance and contact tracker
A startup that created a dedicated gadget to help ensure the safety of electrical industry workers has turned their talents to addressing the need for similar workplace protections in the face of another threat: COVID-19. Vancouver-based Proxxi is launching Halo, a wrist-worn wearable device that can provide a vibration notification to alert someone of the presence of another band within 6 feet – the recommended span of separation to ensure proper social distancing.