You'd think the brilliant minds at Microsoft would have their hands full dreaming up ways virtual reality can be used in real life — or, at least, how to make Windows suck less. Apparently, they also have time to play around with machine-learning algorithms on photos, so they can guess how old you are based on a selfie.
Microsoft's Machine Learning team used the company's Face API (which detects faces in photographs) to create a site, How-old.net, where you can upload a photo of yourself (or anyone, for that matter), and it will guess your gender and age. The results are sometimes accurate...and sometimes not.
I uploaded three photos of myself: One shot in a coffee shop, not smiling, wearing mascara and very light makeup; one #gymselfie with zero makeup, lots of sweat, and a flushed face; and one shot with medium-level makeup, including liquid eyeliner. In the first, it guessed I was 27 (without revealing my exact age, this first guess was the closest); in the eyeliner-shot it estimated 22 (how flattering!); and in the gym shot, I aged all the way to 39. Note to self: A little eyeliner goes a long way?
Daily Dot reporter Selena Larson had a similar experience with Microsoft's site guessing a range of ages for her photos.
AdvertisementADVERTISEMENT
msft thinks i’m 30, but also 25. pic.twitter.com/nyH1ETM6Mw
— Selena Larson (@selenalarson) April 30, 2015
Developer David Barnard tested out How-old.net using one of those self-enhancement apps, and found that the resulting image shaved 13 years off his actual visage, according to the site's algorithmic analysis.
I see why people like those selfie touch up apps, a little “enhancement” took 13 years off my perceived age! pic.twitter.com/WyM9JZgYQN
— David Barnard (@drbarnard) April 30, 2015
But, for some reason or another, the site had issues with Mark Karayan's photo. Despite a full beard, it thought he was 9 years old. Perhaps it's the boyish smile?
@redgirlsays @selenalarson pic.twitter.com/Sq8NWSjdA9
— Mark Karayan (@Mgkarayan) April 30, 2015
It seems that for many people, the exact age this site guesses is wrong — but often, it does get close. Perhaps Microsoft should add a feedback button to let us share how old people in uploaded photos actually are.
How-old.net was actually part of an experiment looking at real-time usage analytics and face analysis. You can give it a spin for yourself. Let us know if it's eerily accurate, or if it thinks your millennial vibe is more baby-boomer. You can share your results with the hashtag #HowOldRobot.
AdvertisementADVERTISEMENT