Facebook launched a deepfake video detection contest with a prize pool of up to $ 10 million

Facebook has announced a public contest to develop technology for detecting deepfake videos. The contest, Deepfake Detection Challenge (DFDC), will have both rankings and prize money.

For this contest, the company will release a data set of faces and videos. This data set will be built by commissioning paid actors, and the company promises not to use any Facebook user data. The competition will be officially launched with the accompanying data set in December at the Conference on Neural Information Systems in Vancouver, Canada.

In the company’s blog post on Thursday, Facebook CTO Mike Schroepfer said the goal of the competition was to ” create technology that everyone can use to better detect when AI is being used.” used to edit a video to mislead viewers . ”

In total, Facebook will spend 10 million USD on this program. Grants and bonuses will be awarded ” to promote the industry as a whole to create new ways to detect and prevent media manipulation through the use of AI to mislead people. different . ” Schroepfer writes.

According to the DFDC website, this competition will be carried out throughout 2020. Winners will be selected by using a ” testing mechanism that allows teams to score the effectiveness of each model, based on confronting one or more test suites from our founding partners . ”

Along with Facebook, there are other partners supporting this competition, including Partnership on AI, Microsoft, and research institutes such as Cornell Tech, MIT, University of Oxford, UC Berkeley, University of Maryland, College Park. , and University of Albany – SUNY.

Mark Zuckerberg appeared in a deepfake video.

Deepfake is becoming a problem for spreading fake news and especially it is even more dangerous with the ability to easily share on a social network with a huge scale like Facebook.

While the social network is still confused with this problem, some deepfake videos created with the faces of Mark Zuckerberg or Senator Nancy Pelosi are released, which forces the company to quickly find public solutions. technology to detect these manipulated videos before they become so subtle they cannot be detected by the naked eye.

Refer to Motherboard