Facebook and Microsoft are co-sponsoring a contest focused on detecting AI-generated "deepfake" videos.
The Deepfake Detection Challenge will run from Oct. to next March with the goal of developing new systems that can identify when a video clip is actually the product of AI algorithms.
Although doctored images are nothing new, the technology behind deepfakes takes the trickery to a whole new level; thanks to today's machine learning programs, you can swap someone's face with another, or take existing footage and manipulate the subject's lips to say something else. The results can produce images and video that look astonishingly real, but are actually fiction.
"People have manipulated images for almost as long as photography has existed. But it's now possible for almost anyone to create and pass off fakes to a mass audience," said Antonio Torralba, a professor at MIT, in a statement. Other algorithms can convincingly clone your voice to make it say whatever the programmer wants. It's why everyone, including US lawmakers, are concerned the technology behind deepfakes risk one day fooling the public.
"Yet the industry doesn't have a great data set or benchmark for detecting them, Facebook's chief technology officer Mike Schroepfer said in today's announcement. "We want to catalyze more research and development in this area and ensure that there are better open source tools to detect deepfakes."
Other co-sponsors of the contest include academics at seven universities, and the industry trade group, Partnership on AI, which will help oversee the challenge.
The contest will work like this: A group of paid actors has been hired to create a dataset of videos. Once recorded, Microsoft, Facebook and the other co-sponsors plan to take the videos and tamper them using the latest AI techniques to generate the deepfakes.
It'll then be up to contest participants to create programs to detect the fabricated videos. The judging will take place by pitting the programs against the deepfakes in a secluded "black box" computer. Facebook, Microsoft and the academics plan on developing a testing mechanism that'll score the effectiveness of the participants' deepfake-sniffing programs.
The contest, which is open to all, plans on offering grants and awards. However, participants will need to go through a screening process, and agree to certain terms on using the video data, as a safeguard against bad actors from accessing the code and data from the contest.
The sponsors will kick off the challenge next month by first testing the quality of the videos they created with the paid actors. Schroepfer said Facebook is investing more than $10 million to fund the effort. The company itself also plans on entering the challenge, but won't be accepting any financial prize if it wins.