Artificial intelligence scans face to measure age

Instagram is testing new ways for users to verify their age, including an artificial intelligence tool developed by third-party company Yoti that can estimate your age by scanning your face.

#According to official regulations, you must be 13 years or older to register an Instagram account. But over the years, the company has made little effort to enforce the rule. It doesn't even bother to ask new users for their birthdays, let alone verify the information. However, it wasn't until 2019 that Instagram rolled out more and more age verification features and ways to distinguish younger users from adult ones, after coming under fire from privacy and child safety experts.

Currently, Instagram will ask for age verification when teenagers try to change their date of birth to show that they are 18 or older. To verify their age, users can send photos of various ID cards. Today, users in the United States will have two additional options: Social Guarantee and AI Assessment.

The first method is social proof, where Instagram will ask the user’s three mutual followers to confirm their age. Mutual followers must be 18 or older and have three days to respond to Instagram's request. The second method, an AI assessment, involves sending a selfie video to Yoti, a third-party company that uses machine learning to estimate a person's age.

Yoti is a well-known company in the field of online identity verification, and its technology has been recognized by the UK government and the German digital regulator. It utilizes various facial signals to estimate the target's age. Although the company itself doesn't know exactly what these signals are.

You can actually try Yoti's system (the company says it doesn't retain any data you share with it) and see how accurate it is below. These numbers show Yoti's error rate in age estimates for different age ranges, skin colors, and genders.

Data shows that Yoti’s system has lower prediction accuracy for female faces and people with darker skin tones, and its estimation error can be as high as 2.5 years for people under 24 years old. However, if the tool only made a rough guess at the user's age, its accuracy would improve. A third-party nonprofit analyzed Yoti's system in 2020 and found that it was 98.89% reliable at guessing whether an 18-year-old was older or younger than 25. (This means that out of 1,000 guesses, the system will only get it wrong 0 to 11 times.) It's unclear how the system will translate to Instagram's use cases, or whether the system will actually be secure enough. For example, you can fool Yoti's web demo simply by holding a photo in front of your camera (though Instagram says this demo doesn't include the anti-spoofing features it's integrating into the official version). Underage users may also seek help from older friends to pass the test.

This isn’t the only AI tool Instagram is using to try and assess a user’s age, either. The company has been using automated systems to search for underage users since 2019. These tools can scan a user’s birthday posts and age in their circle of friends, among other information. For example, if a user claims to be 20 but posts to celebrate their 17th birthday, their account will be flagged and they may need to verify their age. Instagram says it's working to add new data points to the system and is even testing a language analysis tool that can tell whether a user is an adult or a teenager based on how they write.

The above is the detailed content of Artificial intelligence scans face to measure age. For more information, please follow other related articles on the PHP Chinese website!