Facebook announced Tuesday that it will stop using facial recognition technology for identifying people in photos and videos, and erase any associated data on over 1 billion people.
This news marks the end to one of the most popular facial recognition systems. Facebook's auto tag, which is used for face unlocking smartphones and airport applications, is the most commonly encountered facial recognition technology. Jerome Pesenti, Facebook's VP of artificial Intelligence, stated that the decision was made to balance the benefits of facial recognition with growing social concerns.
Since 2010, Facebook has used facial recognition technology to automatically identify people in photos, videos and Memories. This has drawn criticism from privacy advocates as well as hundreds of millions of dollars in fines by government regulators. WIRED was told by a spokesperson for Facebook that labels on billions of photos that were tagged using facial recognition in the last decade will be retained. The signals and cues about a person's social circle, which were gathered using facial recognition photos and videos, will presumably also be preserved.
Face recognition has become a symbol of privacy and human rights, leading more than a dozen US cities to ban the use of this technology. Multiple wrongful arrests have been made in the US due to facial recognition. This has also led to the establishment of a surveillance state in China to manage Muslim minorities.
After weeks of intense scrutiny, Facebook has taken the decision after thousands of internal documents were leaked that showed flaws in moderation. This decision comes after Facebook's last week move to rebrand under the Meta name.
Facebook will continue to use facial recognition technology for photo tags, but it doesn't mean that the company has stopped using facial recognition. Facebook will continue to use facial recognition technology to help users get into locked accounts or verify their identities to process transactions. Facebook will erase data from more than one billion people, but it will keep DeepFace, an AI model that was trained with this data, intact. One in three Facebook users uses the service, which recommends that people tag photos.
Facebook has also removed automatic photo tags. However, Facebook will no more use facial recognition to identify individuals by name in a small number of photos that are intended for visually impaired or blind people.
Facebook is the latest tech giant to discontinue facial recognition. Last year, IBM stopped providing facial recognition services to customers. In the wake of George Floyd's murder, Amazon and Microsoft stopped selling facial recognition services due to a lack of regulatory action.
If this encourages policymakers to consider facial recognition seriously enough, this could be a turning point. Deb Raji is an AI researcher.
Joy Buolamwini and Deb Raji were the first to show that facial recognition systems are less accurate for women with darker skin. These results were confirmed by an analysis of the National Institute of Standards and Technology that found that facial recognition systems often misidentify young Asian people.
Raji, who worked on ethics and policy issues for companies like Google and AI Now, said that Facebook's move was significant because DeepFace had played an important role in the history and evolution of computer vision. In 2014, the deep learning model was built using 4 million images from over 4,000 people. This is the largest ever dataset of faces. DeepFace was the first AI-based model that could recognize facial expressions at human-level. It sparked a new trend in commercializing the technology and hoarding data for performance improvement.