Twenty Thousand Hertz Latest Episode: Exposing Audio Deep Fakes

 Last week, a tweet from Rep. Steve Scalise (R-LA) showing a video of activist Ady Barkan received a “manipulated media” label from Twitter. Barkan has ALS and speaks through voice assistance.

In the video, a conversation between Barkan and Democratic presidential nominee Joe Biden, Barkan asks “But do we agree that we can redirect some of the funding?” The version Scalise tweeted edits in the words “for police,” to the end of the question, words which Barkan says in a different context earlier in the video.

Last year, a fake video made the social media rounds purportedly showing Speaker of the House Nancy Pelosi appearing to be drunk. The video was slowed down to make it appear the Pelosi was impaired. In August, Fox News photoshopped Donald Trump out of a photo with Jeffrey Epstein. 

In late July, Georgia Democratic Senate candidate Jon Ossoff announced that a campaign ad by Senator David Perdue (R-GA) showed a manipulated image with Ossoff's nose obviously elongated to make him look more ethnic.

Between the ubiquity of Photoshop and the Internet’s love for videos that deceptively depict one person’s face on another person’s body, it’s easier than ever to not believe every incriminating or humiliating image you see.

However, we are entering a brand new era and on today’s episode of TED podcast Twenty Thousand Hertz, host Dallas Taylor explains why we can no longer trust our eyes or ears: welcome to the world of audio deepfakes, where Joe Biden covering CeeLo Green’s “Forget You” is the least of our post-truth, hyper-partisan concerns.

Listen to Deepfake Dallas, featuring a cameo from “George W. Bush”: https://www.20k.org/episodes/deepfakedallas

Co-hosting this episode with a deepfake version of himself, Dallas investigates how Artificial Intelligence and modern machine learning have given us the highly accessible and increasingly indistinguishable ability to create fraudulent versions of people’s voices, saying whatever one wants them to.

The possibilities for memes are endless, but so are the implications for much scarier societal, political impacts. After breaking down the technological process with deepfake wizard Tim McSmythurs, Dallas welcomes Riana Pfefferkorn, Associate Director of Surveillance and Cybersecurity at Stanford Law School, to analyze the legality of impersonations, whether audio deepfakes could actually be a form of protected speech, and how funny clips could evolve into financial phishing scams, soundbites that spread racism and sexism, false declarations of military invasions, martial law and more, influencing elections, national security and personal safety.  

 

 Twenty Thousand Hertz began as a podcast in 2016 with its first episode exploring the voice of Siri. In early August, Twenty Thousand Hertz became part of a new family of podcasts being launched by TED. TED is a nonprofit devoted to spreading ideas, usually in the form of short, powerful talks (18 minutes or less). Popular TED podcasts include The TED Radio Hour, Worklife With Adam Grant and TED Interview.

 

 
 

The podcast also celebrated its 100th episode by dissecting the sonic brand of Netflix. It includes interviews with the company and a breakdown of a familiar audio logo to millions of Netflix subscribers. The Netflix sonic signature consists of those two bass-heavy hits you hear when you open the app or watch any of their content. 

Listen to the episode here: