Soon, Alexa Will Know When You’re About to Die

Smart speakers and other home technologies continue to raise serious privacy and security concerns. Many of us own smart devices because the convenience outweighs the negative, indirect effects they have on our lives and society as a whole. While these data collection practices typically come with complex ethical issues that most of us still struggle to resolve in our daily lives, it’s important to remember that the practices, themselves, are morally agnostic. However, researchers at the University of Washington have discovered that continuous data collection from your smart speaker could save your life.
It’s not a stretch to imagine a scenario in which a person in an emergency situation asks Alexa to get help and it utilizes its numerous connections to get help. In fact, you don’t have to imagine it because it has already happened–just not on purpose. Regulatory issues prevent Amazon (and other companies) from adding emergency features easily but it’s far from a technological limitation. In fact, Alexa seems to understand how to facilitate a 911 call to the confusion of all involved (even though it’s obviously because of a Bluetooth phone connection and not because Alexa became sentient).
But what happens when you can’t say the command to get help in the first place? You can’t always voice a command in an emergency because you’re physically unable or doing so would put you in more danger. Smart devices may offer another way to help, beginning with cardiac arrest and the way you breathe. The initial signs of a heart attack include irregular gasps of breath, formally known as agonal respiration, which provide a unique sonic pattern a machine can identify—so long as it’s always listening. Bloomberg spoke to Dr. Jacob Sunshine, an assistant professor of anesthesiology and pain medicine at the University of Washington, to get the details:
This kind of breathing happens when a patient experiences really low oxygen levels. It’s sort of a guttural gasping noise, and its uniqueness makes it a good audio biomarker to use to identify if someone is experiencing a cardiac arrest.
The researchers trained their initial AI model using 7,316 2.5-second audio clips that were captured by smart devices and cellphones over the course of eight years. They also used around 83 hours of normal sleeping and breathing sounds to provide negative data in order to reduce the misidentification of agonal respiration in the model. Their system was able to accurately detect agonal breathing events up to six meters away and act.
While the system can easily run on systems with hardware in existing smart speakers like the Amazon Echo, and efforts have been made to fine-tune accuracy, the researchers believe their model needs further training before it can function reliably in real-world conditions. That’s a good thing, too, because if it doesn’t work properly it’s just invading your privacy without helping you. We have enough of that already.
Top image credit: Adam Dachis
Continue reading

Scientists Find Planet Where It Rains Molten Rock
The ground is rock, the seas are rock, and yes, even the air is rock.

What Does It Mean for the PC Market If Apple Makes the Fastest CPU?
Apple's M1 SoC could have a profound impact on the PC market. After 25 years, x86 may no longer be the highest-performing CPU architecture you can practically buy.

Which Is Faster, the Xbox Series X or PlayStation 5? Early Data Says It’s Complicated
Competitive head-to-head data on the Xbox Series X versus the PlayStation 5 is beginning to trickle out.

Google Kills Free Photo Storage, Changes What Counts Toward Storage Caps
Google has announced some significant changes to Photos, especially if you use the service for automatic backup.