Face recognition algorithms aren't as reliable as claimed, study finds

Artificial intelligence can spot your face in a crowd of thousands with near-perfect accuracy, surpassing the ability of humans to do the same – but when confronted with a larger set of images, it's not quite up to par. In the MegaFace Challenge launched by the University of Washington, researchers are working to improve the capabilities of facial recognition algorithms at the million person scale. It's hoped that the competition will help to solve crucial problems in facial recognition, including the identification of a single person across different ages and poses. Artificial intelligence can spot your face in a crowd of thousands with near-perfect accuracy, surpassing the ability of humans to do the same – but when confronted with a larger set of images, it's not quite up to par Recently, facial recognition algorithms have proven their abilities to perform with near-perfect accuracy. But, these algorithms were trained on a dataset of just 13,000 images, and when confronted with a larger collection, accuracy dropped.