Episode 69: Demystifying Deepfakes

Generative artificial intelligence has captured the world’s attention after recent advances in the commercial sector. Its ability to create deepfakes, or highly realistic multimedia, has turned a once highly specialized skill into something as easy as clicking a button. As a result, the threat of manipulated media –audio, images, video, and text – has increased while social media provides a ripe environment for viral content sharing. Though, not all media manipulations have the same real-world impact. In this episode of Voices from DARPA, Dr. Wil Corvey, program manager for DARPA’s Semantic Forensics (https://www.darpa.mil/program/semantic-forensics) (SemaFor), discusses how the program goes beyond detection to delve deeper into understanding the intent behind manipulated media and how their team is creating tools available for today’s analysts. They are joined by SemaFor researchers Arslan Basharat, assistant director of Computer Vision at Kitware Inc. and Luisa Verdoliva, professor at the University of Naples in Italy.

Om Podcasten

DARPA’s podcast series, "Voices from DARPA," offers a revealing and informative window on the minds of the Agency's program managers. In each episode, a program manager from one of DARPA’s six technical offices—Biological Technologies, Defense Sciences, Information Innovation, Microsystems Technology, Strategic Technology, and Tactical Technology—will discuss in informal and personal terms why they are at DARPA and what they are up to. The goal of "Voices from DARPA" is to share with listeners some of the institutional know-how, vision, process, and history that together make the “secret sauce” DARPA has been adding to the Nation’s innovation ecosystem for nearly 60 years. On another level, we at DARPA just wanted to share the pleasure we all have every day—in the elevator, in the halls, in our meeting rooms—as we learn from each other and swap ideas and strive to change what’s possible.