Advancements in computer vision and deep learning have gained momentum in the last few decades due to the utilization of graphics processing units in training neural networks. While there are a plethora of beneficial use-cases of these highly accurate algorithms, researchers and developers alike have not turned a blind eye to their ignoble corollaries. One massive disadvantage that they are worried about is that malicious actors can utilize these systems to spread disinformation by producing "deepfakes", which are realistic AI-generated videos of real people saying and doing fictional things.
To combat this, Facebook announced today the Deepfake Detection Challenge (DFDC) in collaboration with the Partnership on AI, Microsoft, and academics from Cornell Tech, MIT, University of Oxford, UC Berkeley, University of Maryland, College Park, and University at Albany-SUNY. The goal of this challenge will be to create a substantial dataset that can be used to train models that accurately detect deepfakes and report them.
The social network giant is donating over $10 million alone to fund this effort and hopes that through this initiative, it will incentivize researchers and organizations to come up with novel ideas to combat the nuisance of disinformation on social media and the world wide web in general. The Chief Technology Officer at Facebook, Mike Schroepfer, wrote:
We want to catalyze more research and development in this area and ensure that there are better open source tools to detect deepfakes. [...] The Deepfake Detection Challenge will include a data set and leaderboard, as well as grants and awards, to spur the industry to create new ways of detecting and preventing media manipulated via AI from being used to mislead others.
The efficacy of the dataset and the models trained with it will be gauged at the International Conference on Computer Vision (ICCV) through a technical working session. The resultant, full dataset will be unveiled in December, at the Conference on Neural Information Processing Systems (NeurIPS).
Furthermore, Facebook has clarified that it will not be using any user data for the purpose of creating this dataset and that only consenting and paying actors will be able to take part in the challenge. For more details on the official announcement, you can read the blog post here. For more information on the challenge, visit the Deepfake Detection Challenge website here.