This AI predicts how old children will be. Can It Keep It Safe?


Guess how old a person based solely on their appearance is extremely difficult to heal, especially those awkward early in adolescence. And yet bouncers, liquor store owners, and other keepers of age-restricted items make that quick estimate all the time.

Their predictions are often wrong. Now London -based digital identity company Yoti believes that AI-driven age estimation can predict how old a person will be if they are aged anywhere from 6 to 60. For the first time, it says, it is accurate to determine if children are under or over 13, the minimum age of many social media Companies need their users.

Yoti’s image technology could be even more attractive as faced by Big Tech and internet services increased scrutiny how children use their products. However, privacy advocates say that automatic analysis of people’s faces normalizes monitoring, is largely unregulated, and has the potential to show bias.

Yoti says its age-estimation technology, which it has developed over the past three years, has a margin of error of 2.79 years across the entire 45-year age range. For those under 25s the margin of error is reduced to less than 1.5 years. In the coming weeks, it will take brick and mortar tests at five major supermarket chains in the UK. The company did not name the supermarket brands, but said several unnamed pornography and gaming websites were also testing the technology. stop minor visitors. It added that age estimation technology is already being used for streaming on the children’s social network Yubo and the healthy live app Smash.

Point a camera running Yoti’s software at your face — it can be accessed via the web on your phone, laptop, or tablet, or at the self-checkout terminal — and the system estimates your age. In many tests using a browser-based game environment on my phone, the system correctly placed me on the 27–31 and 28–32 ranks. The company states that it or its clients will not store the images it captures, and you do not need to register to use them. “It’s not recognition. It does not certify any individual, ”said Julie Dawson, director of regulation and policy at Yoti. The company claims it does not recognize face, because it does not recognize individuals. “If it sees a new face, it will just release the estimated age of that individual,” Dawson said.

Yoti clients can also use criteria for estimating age: for example, setting a estimate limit of 25 if someone in the UK buys alcohol they must be over 18 by law. Anyone flagged below that threshold may be asked to provide ID. The system also informs its customers how much it trusts any given estimate.

The company trained its neural networks with “hundreds of thousands” of photos of people’s faces, according to Yoti cofounder and CEO Robin Tombs. It usually collects the faces themselves, through its standalone Yoti app that allows people to verify their ID with governments and other bodies by uploading official documents such as passports and driver’s licenses. If people upload their details to the Yoti app, they have the option to opt out of the data used in Yoti’s AI training. The company itself isn’t sure what facial features its AI uses to determine people’s age. “We have to be honest,” Tombs said, “we really don’t know if it has anything to do with wrinkles or saggy eyes or whatever. It’s done so much now that it’s so good at it. ”



Source link

admin

Leave a Reply

Your email address will not be published. Required fields are marked *