August 14, 2019

Why Are Deepfakes So Effective?

It's because we often want them to be true.

Public opinion shifts, skewed election results, mass confusion, ethnic violence, war. All of these events could easily be triggered by deep fakes—realistic seeming but falsified audio and video made with AI techniques. Leaders in government and industry, and the public at large are justifiably alarmed. Fueled by advances in AI and spread over the tentacles of social media, deep fakes may prove to be among the most destabilizing of forces humankind has faced in generations.

It will soon be impossible to tell by the naked eye or ear whether a video or audio clip is authentic. While propaganda is nothing new, the visceral immediacy of voice and image give deep fakes unprecedented impact and authority; as a result, both governments and industry are scrambling to develop ways to reliably detect them. Silicon Valley startup Amber, for example, is working on ways to detect even the most sophisticated altered video. You can imagine a day when we can verify the authenticity and provenance of a video by way of a digital watermark.

Developing deep fake detection technology is important, but it's only part of the solution. It is the human factor—weaknesses in our human psychology—not their technical sophistication that make deep fakes so effective. New research hints at how foundational the problem is.

After showing over 3,000 adults fake images accompanied by fabricated text, a group of researchers reached two conclusions. First, the more online experience and familiarity with digital photography one had, the more skeptical the person evaluating the information was. Second, confirmation bias—the tendency to frame new information to support our pre-existing beliefs—was a big factor in how people judged the veracity of the fake information.

Read the full article in Scientific American.

  • Commentary
    • Just Security
    • September 19, 2024
    Competition, Not Control, is Key to Winning the Global AI Race

    The United States, with much of the world’s AI-enabling infrastructure, has positioned itself as the global leader in AI innovation. That might not be the case for much longer...

    By Keegan McBride & Matthew Mittelsteadt

  • Commentary
    • Time
    • September 16, 2024
    Regulating AI Is Easier Than You Think

    Countries can regulate AI from the ground up by controlling access to highly specialized chips...

    By Paul Scharre

  • Commentary
    • Sharper
    • September 11, 2024
    Sharper: Drones on the Battlefield

    From the battlefields of Libya to Nagorno-Karabakh to Ukraine, the deployment of drones has become a critical element of modern warfare. Will the explosion of unmanned aerial ...

    By Anna Pederson & Molly Campbell

  • Commentary
    • Lawfare
    • September 5, 2024
    How to Revamp Chinese Students’ American Education

    The PRC today operates the largest and most sophisticated propaganda apparatus in human history....

    By Bill Drexel & Grace Gao

View All Reports View All Articles & Multimedia