YouTube Introduces Age Prediction Feature Through AI Technology
YouTube has announced the trial of a new age verification system, utilising machine learning to detect underage users. This innovative approach marks a departure from traditional methods like facial recognition and is part of the video-sharing platform's broader efforts to protect younger viewers.
The system, currently in a trial phase, is limited to some U.S. users. Unlike traditional methods, it does not ask for birthdates but instead studies user behaviour, such as video searches, watched categories, and account age.
If the system flags a user as a minor, it activates safety mode. In this mode, personalised ads, access to digital wellbeing tools, and recommendations to prevent binge-watching potentially harmful content are restricted. However, it's important to note that the system does not reveal whether it has classified a user as a minor.
YouTube's machine learning-based age estimation system aims to overcome the physical limitations of traditional methods, which often show mean errors of about six years and have difficulty accurately identifying younger users. By focusing on digital behaviour, the platform hopes to provide a more age-appropriate experience, especially for teen users.
However, this behavioural AI system is not infallible. It can sometimes incorrectly classify an adult user as a minor, requiring users to verify their age via government ID, credit card, or selfies if flagged incorrectly.
The major challenges include potential biases and inaccuracies. Security researchers warn these AI age verification systems can disproportionately misclassify minorities and people from lower socioeconomic backgrounds due to limitations in AI models and hardware infrastructure. Users and advocates have voiced concerns over privacy and the security of submitting sensitive government IDs to Google.
In terms of effectiveness, YouTube's system does not disclose how it determines the likelihood of a user being under 18 in detail. This lack of transparency raises questions about the system's accuracy and reliability.
It's worth noting that this age verification measure is part of a larger, industry-wide effort to protect children online. Unlike Meta's similar AI tool, YouTube's system does not allow users to easily update their settings to correct mistakes.
In summary, YouTube’s machine learning approach to age estimation represents an innovative shift from physical appearance-based methods to behavioural analysis. While it may improve detection accuracy, it still faces significant technical, ethical, and privacy challenges. The system is a supplement to, rather than a replacement for, age verification and is part of broader efforts to comply with stricter global age verification laws.
[1] BBC News: YouTube trials AI to spot underage users [2] The Verge: YouTube is testing a new AI system to detect underage users [3] TechCrunch: YouTube is testing AI to detect underage users in the US [4] Wired: YouTube's AI-Powered Age Verification System Has Privacy and Bias Concerns
- The trial of YouTube's new age verification system involves artificial-intelligence, studying user behavior such as video searches, watched categories, and account age to detect underage users.
- Despite its innovative approach, YouTube's machine learning-based age estimation system faces challenges, particularly potential biases and inaccuracies that may disproportionately affect minorities and people from lower socioeconomic backgrounds.
