How D-ID is using AI to change facial recognition
This article is part of our “How AI is changing the world” event series, held in San Francisco, New York, and Tel Aviv from June to November 2019, featuring insights by leading scientists and entrepreneurs on how AI will change healthcare, communication, agriculture, travel, and other industries. Check out all 12 talks here.
******
At D-ID, founder and CEO Gil Perry is working to keep online identities — specifically, people's faces — a secret. "Right now, we are losing the basic human right to privacy. This is because of facial recognition," Gil said. The amount of visual data stored by corporations is growing in general, and personally identifiable faces are no exception. This problem is common in self-driving car datasets which can record many pedestrians on each trip.
At D-ID, data scientists developed AI software that can change photos to make them unrecognizable to facial recognition software, while indistinguishable to humans. The technique is based on a proprietary process that iteratively modifies the face in a photo to maximize the facial dissimilarity in an ensemble of prominent facial recognition models, while minimizing the dissimilarity in a human visual model.
D-ID performs a similar task for videos, replacing faces with computer avatars that still retain a person's facial expression, gaze, emotions, age, and ethnicity. Both technologies will help corporations comply with emerging regulations, which prohibit them from publishing personally identifiable information, such as a person's face. "We are filling the gap between regulation, technology, and reality," said Gil.
This video avatar replacement allows for personalized ads where in the past regulation would prevent it. After being processed by D-ID, advertising can be based on non-personally identifiable attributes that are preserved in the video by this processing technique. Perhaps more importantly, the services provided by D-ID might be able to restore some of the privacy we’ve already lost.