Everything in Your Archive
is Now Fake
Presented at the annual conference of the
Association of Moving Image Archivists
Friday, November 30 2018
“It’s been a little bit of a fluke, historically, that we’re able to rely on videos as evidence that something really happened.” - Ian Goodfellow, Google Brain
In early 2018 users of a simple new program, harnessing advances in AI and graphics processing, began creating alarming, inexpensively-produced and eerily-realistic pornography of celebrities- created by face-swapping their likenesses onto pornographic actors. This was the dawn of “deepfakes.”
Today, with just a pool of pictures and a click of a mouse, it is easier than ever to concoct increasingly believable deepfake videos of anyone. Deepfakes sow doubt and undermine the credibility of all moving images. New deepfakes will include revenge-video and fake news of politicians, law enforcement, and the military, but also hybrid-reality experiences with actors and avatars both real and not, living or dead. This session will introduce deepfakes, show how they’re made, recent examples, and discuss both their impact and opportunity.
Deepfake News Bibliography
News Sources April 2018 - June 2019
(no longer updated- maintained by reader request)
(scroll down for more high-quality deepfakes)
about the speakers
Gaurav Oberoi is an Entrepreneur in Residence at the Allen Institute for AI, where he is exploring commercial applications for the latest academic research in AI.
His interest in generating hyper-realistic product photos led him to explore the Deepfakes phenomenon, and publish a detailed essay and open source code that was widely read. Gaurav will share his findings with us in the session titled “Everything in Your Archive is Now Fake.” In the past, Gaurav has started and sold two startups (BillMonk, and Precision Polling), worked as an engineer at a large tech giant (Amazon), and founded the Audience business at SurveyMonkey as it grew from 50 to 700 people.
"Exploring DeepFakes" Medium, March 5, 2018
Yvonne Ng is an audiovisual archivist and has been part of the WITNESS team since 2009. In collaboration with WITNESS regional leads, she trains and supports partners on collecting, managing, and preserving video documentation for human rights advocacy and evidence. She develops training resources related to archiving and preservation, such as the groundbreaking Activists’ Guide to Archiving Video. Yvonne also manages WITNESS’s own archive of human rights video.
Outside of WITNESS, Yvonne currently serves on the Board of Directors for the Association of Moving Image Archivists (AMIA), on the Advisory Boards of the Memory Lab Network and Documenting the Now, and as a Research Consultant for the Human Rights Investigations Lab at UC Berkeley. She regularly participates in international archiving fora and tweets on archival issues at @ng_yvonne.
Prior to joining WITNESS, Yvonne worked as a Research Fellow on the Preserving Digital Public Television Project, and at NYU Libraries, New York Public Library, and the Canadian Filmmakers’ Distribution Centre. Yvonne holds an MA in Moving Image Archiving and Preservation from New York University, where she has also taught a course on Personal Digital Archiving. She holds a BA in Cinema Studies from the University of Toronto.
John Tariot of Film Video Digital works both in the past: preserving motion picture film for a national clientele of archives and film and TV production professionals, and in the future: developing digital strategies for motion picture archives.
A futurist with a track-record, John led the advent of the online stock footage economy with the creation of FOOTAGE.net, which included designing and building groundbreaking online databases for NBC News, National Geographic, Paramount Pictures, CNN, and many others.
Now, current work is focused on AI’S coming impact on motion picture archiving, some of which will be shared in this session. John's career also includes serving on the Advisory Committee for Peter Gabriel's human rights archive WITNESS- and on the Advisory Board for the broadcast-industry bible POST Magazine. He is a member of the Association of Moving Image Archivists (AMIA), the American Film Institute (AFI), and his expertise in color correction and image restoration has been recognized by acceptance to Colorist Society International (CSI).
Follow-up interview with Hany Farid
John Tariot interviews Hany Farid, professor at Berkeley School of Information, where he focusses on digital forensics, image analysis, and human perception. He is one of the subjects in the New Yorker article, “In the Age of A.I., Is Seeing Still Believing?” Hany and John Tariot have been having an ongoing conversation about the impact deepfakes will have on archives, and he and John continue the discussion here on some of the issues raised at the Association of Moving Image Archivists conference session “Everything in Your Archive is Now Fake.”