“Deepfake” methods, which current realistic AI-generated movies of actual people saying and doing fictional pieces, hold important implications for deciding the validity of the information presented on the internet.
The”Deepfake Detection Challenge” will incorporate a data collection and leaderboard, in addition to awards and grants, to spur the industry to make new methods of discovering and preventing media controlled via AI from being used to deceive others.
“No Facebook user information will be utilized within this data collection. We’re also funding research collaborations and prizes to the struggle to help promote more involvement. In total, we’re dedicating over $10 million to finance this industry-wide campaign,” Schroepfer said in a statement.
The entire data set to release along with also the DFDC launching will occur in the Seminar on Neural Information Processing Systems (NeurIPS) that December.
“To be able to move in the information age to the knowledge era, we have to do better at identifying the real from the fake, benefit reliable content above untrusted articles, and instruct the next generation to become better electronic taxpayers,” explained Professor Hany Farid from UC Berkeley.
“The objective of the contest is to build AI systems which could detect the small imperfections at a doctored picture and expose its fraudulent representation of fact,” additional Antonio Torralba, director of the MIT Quest for Intelligence.
“Deepfakes” are movie forgeries which make individuals seem to be saying they never did, such as the favorite forged videos of Facebook CEO Mark Zuckerberg and US House Speaker Nancy Pelosi that moved viral lately.
“Given the current advancements in having the capability to create manipulated data (text, pictures, videos, and sound ) at scale, we all need the entire participation of the research community within an open environment to come up with systems and methods which could detect and mitigate the ill effects of exploited multimedia,” noted Professor Rama Chellappa in the University of Maryland.