Facebook and Microsoft launch deepfake detection contest

Facebook, together with Microsoft and the Coalition for Partnerships for Artificial Intelligence for the Benefit of People and Society (PAI), as well as scientists from several universities, announced a competition for the development of deepfake recognition technology.   According to a Facebook blog, as part of a project called the Deepfake Detection Challenge, Facebook will instruct researchers to create realistic diphakes. After that, the contestants will work out deepfake detection algorithms. Authors of the best will receive grants. The prize fund of the project is $ 10 million.   The Deepfake Detection Challenge starts in October, and will end in March next year. The first diplomas for the contest will be tested at the International Conference on Computer Vision, and the entire set of videos for work will be released in December.   “The industry does not have enough data sets or benchmarks to detect them. We want to stimulate more research and development in this area and develop better open source tools for detecting fakes, ”Facebook said.   Deepfake (videos created by a neural network) has long attracted the attention of American authorities. In August, the University of Oregon announced a study in which scientists will try to teach mice to recognize artificial speech. The Department of Defense Advanced Research Projects Agency (DARPA) over the past two years has spent $ 68 million on developments designed to detect diphills. Since 2016, the agency has been working with the Media Forensics (MediFor) team, which is engaged in development in this area.   Technology itself, meanwhile, is becoming more accessible. Last week, a Chinese application called Zao, which allows users to turn their faces into movie star faces, reached the top in the country’s app store. The service was launched on August 30 and quickly gained popularity in two days. The number of references to it in the Chinese counterpart of Twitter (Weibo) exceeded 8 million. True, the application almost immediately ran into a negative reaction. China’s main social network, WeChat, has even blocked the service.