Voice assistants, known formally in the tech industry as artificial intelligence or AI, are useful for all sorts of things, but a new study says that AIs such as Amazon Alexa will have future capabilities such as predicting the onset of heart attacks and calling for help.
The study, conducted by the University of Washington and mentioned in the journal Digital Medicine, uses the microphone in a smart speaker such as Amazon's Echo lineup or a smartphone to listen for the gasping sound characteristic of oncoming heart attack and then calls for help. The result of artificial intelligence in situations such as a sudden heart attack onset could lead to higher survival rates.
To test out this AI capability, researchers trained the system with audio clips from 729 emergency calls in King County, Washington, a total of 82 hours of audio recordings, and paired them with sleep sounds from 35 volunteers and 12 patients participating in a sleep study who suffered from snoring and sleep apnea. Sleep study patients exhibited similar noises to agonal breathing (heart attack) patients, helping to complicate the study so as to ensure the AI and smart speaker could correctly identify genuine emergency patients.
Head researcher Justin Chan says that the results show a 0.2% false positive rate among volunteers and a 0.1% false positive rate in the sleep study. In other words, Amazon Alexa demonstrated that she can rather accurately distinguish between sleep sounds and heart attack breathing with little error. Alexa was accurate in 97% of the cases in the study, even at 20 feet away.
While Chan is positive about the results in the study, he acknowledges that there's some progress to be made and hurdles to be overcome before the system goes live. For example, Chan envisions that there'll be a 15-second to 30-second warning given to patients so that Alexa (or whatever AI you wish) doesn't make a false call to EMS and arrange a pickup due to a false positive. While Alexa proved reliable in 97% of the cases used, there's still some room for improvement and refinement.
Next, there is the major issue of privacy, which has become one of the big drawbacks to smart speakers and AI in general. For Alexa or any other AI (be it Google's Google Assistant, Apple's Siri, or even Microsoft's Cortana) to perceive a heart attack due to user breathing, it would have to be always on, listening for user breathing patterns. Current AIs can be enabled to be always on, but the medical need might necessitate a function that, up until now, has been a matter of preference for individuals.
Those who choose to enable AI and keep it always on may find the AI recording other instances where no heart attack is present, such as in a private bedroom conversation between a husband and wife. The new system would be designed for bedrooms, making the smart speaker or smartphone a problem with regard to privacy.
Amazon's Alexa has come under fire for privacy invasion. Last May, Alexa recorded a woman's conversation with her husband and emailed it to one of the husband's employees. Last March, Alexa suffered from a glitch that caused her to burst out laughing randomly.
And, in one of the creepiest weird happenings ever, Alexa randomly said, in a calm voice, "Every time I close my eyes, all I see is people dying," one evening when a man was walking through his home with his TV paused. Apart from the weird AI glitches that are known to occur, there's also the random activation of AIs even when users do not use the given command ("Hey Google" or "Ok Google") to activate them.
With all these issues surrounding smart speakers and AIs, more testing time is needed. And yet, for all the creepy and weird things AIs can do, if Alexa is any indication, they just might save a life.