It was first invented for Android, later picked up and refined by Apple, and now it's spread to almost all smart devices. Should you have the bad fortune to be in an auto accident, your phone or watch can sense the event and reach out to emergency services. Using inputs from the device's gyroscope, accelerometer, microphone and navigation systems, the feature determines that it (and the person it is attached to) has been in a crash. After flagging it on the screen and giving the owner time to stand it down, it will automatically dial 911 to get you help.
While it seems the stuff of a marketing campaign, there are actually numerous anecdotal examples of the system at work. Wrote one poster "Tire blew out and I hit a telephone pole going about 70. Head against the wheel, continuous honking, smoke coming for the car. Felt the little haptic that authorities were being called. This was in the middle of nowhere so if it weren't for the watch, who knows how long I would've been there. Next thing I know I'm being hauled away in the ambulance." Another couple's car slipped off a gravel shoulder, plunging trough some trees and winding up upside down. When they crawled out their phone had already called for help, and rescuers were on their way.
Impressive to be sure, but there have also been a large number of false positives reported, with crews responding to non-existent emergencies. Extreme sports enthusiasts, with their quick stops and starts, sudden changes in direction and frequent yelling seem to trigger the system, as do roller coasters and skiing. Even the phone falling during your weekend bike ride can cause the stimuli needed to initiate the sequence. Once started, it's like a scene from any "Mission: Impossible" movie where you have to disarm the countdown before it blows up, or this case, dials for help.
Just as we're told it has the potential to change everything else, AI advances should help reduce those miscues. Using a larger data set and computational ability, the hope is that the systems will be better able to differentiate between a real crash and a ride on the Anaconda at Kings Dominion. But the same type of sensory input, large language modeling and ability to synthesize a possible outcome offers the tantalizing possibility of other "detectors" which could be useful in less life-threatening situations. All we need is some 15-year old coder to stay up late and crank these potential million-dollar ideas out. To wit:
Repeat Story Detector. Much like how the "Shazam" app is able to listen to any song and identify it from its melody, beat and lyrics, this feature monitors incoming audio, like stories from your friends. When it detects a single voice talking for more than a preset period of time with only bored "uh hmms" from the device owner's own voice (as likely acknowledgment that it's been heard before), it triggers its Large Language Data Set function to compare that stream of words with all the others it has heard from that same voice. If it finds a match, it initiates a self-call, enabling the device owner to politely interrupt and say, "Whoa... great story, but I have to take this!"
Spoiler Detector. You must first input your favorite streaming shows. The app then uses its audio capabilities to listen and track your viewing habits, keeping a record of how far along you've gotten in the series. The detector then reads ahead in any emails, texts, articles or posts you call up. If it contains a spoiler with the identity of Lady Whistledown's revealer, or what happens to the White Worm once they get to Dragonstone, it sounds an alarm and blacks out the offending passage.
Amazon Equity Detector. This app uses AI to keep scan all delivery services that have your address, including USPS, UPS, and FedEx, as well as Amazon's own system of independent operators. It tabulates all deliveries directed to your home, and aggregates them by addressee. If your total exceeds that of your spouse or partner for a given week, it automatically puts a hold on a variable number of packages while also alerting you so you both receive an equal amount, resulting in fewer accusations of overshopping.
Can't Carry A Tune Detector. Singing along is fine, unless the person singing can't carry a tune in a basket. This app can be set to tolerances ranging from "Slightly Off Key to "Caterwauling." It compares voices singing nearby to whatever MP3 file is playing. If it is not the owner's voice, and it is outside the selected tolerance, it begins to skip and produces a false "WiFi Lost" message. It then changes to a feed of the local NPR station. Once it no longer detects the trigger voice in the vicinity, it returns to playing music.
There's nothing wrong with technology jumping in when a catastrophe strikes and summoning help. But our everyday crises are also ripe for problem solving, and potentially more useful. It's a matter of focusing on what we really want and need, as opposed to what some developer thinks we need. Or as author Joanna Maciejewska put it, "I want AI to do my laundry and dishes so that I can do art and writing, not for AI to do my art and writing so that I can do my laundry and dishes."
-END-
Marc Wollin of Bedford is not sure what half the features on his phone are for. His column appears weekly via email and online http://www.glancingaskance.blogspot.com/ and https://marcwollin.substack.com/, as well as via Facebook, LinkedIn and Twitter.
No comments:
Post a Comment