On September 5, Facebook and Microsoft announced the Deepfake Detection Challenge (DFDC) to produce technology that can be used to detect an “deepfake” videos, that is, AI-generated videos of real people and events that have been altered to mislead. When will the challenge be launched? In December 2019 at the Conference on Neural Information Processing Systems (NeurIPS) in Vancouver, Canada, but events will begin in October 2019. Who built this challenge? Facebook, Microsoft, the Partnership on AI, and academics from Cornell Tech, MIT, University of Oxford, UC Berkeley, University of Maryland, College Park, and University at Albany-SUNY What does the challenge include? Facebook is contributing $10 million to the effort. The challenge includes: Data set of videos and deepfakes that can be used by the community to develop technology around it will be made available to participants. Facebook has clarified that it is working with a third-party vendor to create a new data set of videos using “paid actors, with the required consent obtained”. Facebook will then create “tampered videos” on a subset of these videos using AI. No Facebook user data will be used in this data set. Funding for research collaborations and prizes to encourage participation How will the data set and challenge parameters be tested? Targeted technical working session in October 2019 at the International Conference on Computer Vision (ICCV) Full data set will be released in December 2019 at NeurIPS Who will run it? The Partnership on AI’s new Steering Committee on AI and Media Integrity,…
