A London police department tweeted over the weekend that they had made several arrests due to the presence of “a large amount of cannabis,” along with a photo of what looked like some truly foul weed. Jokes about the police seizing hedge clippings, loose-leaf chamomile tea, and so on flooded the comments, but one tweet featured a fabricated audio clip of Vice President Joe Biden defending his preference for subpar cannabis.
“I’m from Scranton,” the simulated Biden said. “What I’m smoking is dirt.” So let’s get that straight, Jack. Pure brick. Ass. Okay?”
True Biden has never made such a claim (at least not publicly). Digital Biden, on the other hand, has been talking about a wide range of issues where the president neither would nor should take a firm stance. The breakthrough is the result of the recent release of cutting-edge artificial intelligence (AI) tools for voice cloning, specifically Voice Lab, a platform developed by the AI startup ElevenLabs.
ElevenLabs’ audio production software is available to you right now for just $5. Recordings of Biden’s voice are plentiful, so it’s simple to build your AI copy of him. All it needs is a short sample of speech to create a “clone,” and then it can realistically mimic any person and read any text in their voice. After that, just type in his lines of dialogue. Maybe in your head, Biden is a gamer who constantly complains about the “lame-ass bitch boys” who play the “no-skill heroes” in Overwatch. Joe Rizzler can now be a real person in your life.
Although examples like these are obviously fabricated and might at first seem harmless, the potential for abuse of such technology is cause for concern. Shortly after Voice Lab went into beta, ElevenLabs tweeted that the company had seen “an increasing number of voice cloning misuse cases” and would be taking steps to “implement additional safeguards.” While the company did not provide details about the misuse, 4chan trolls were using the tool around this time to generate fake audio of various public figures spewing hate speech.
During one brief section, it sounded like actress Emma Watson was reading from Hitler’s Mein Kampf. Biden seemed to say bad things about transgender people in a deep fake and edited video from earlier this month. That, obviously, did not occur.
Requests for comment from ElevenLabs regarding the security enhancements currently in place or planned for Voice Lab were not met with a response. They also didn’t say whether they consider the non-offensive AI Biden speeches, like the one shared by @MNateShyamalan in which he waxes philosophical about the television show Young Sheldon, to be “misuse.” We don’t yet know what effect voice cloning will have in the long run, but as technology advances, it will make it harder to tell real recordings from those made to promote a false narrative or agenda. At the moment, however, shitposters are enjoying the opportunity to put politically neutral words in the president’s mouth.
The appeal of a Biden bot spouting drug and video game references begs the question: why is this so entertaining, enticing, or even comforting? To begin, it’s not exactly the White House’s style. Plus, Biden is 80 years old; the language of today sounds ridiculous to his grandfatherly ear. Digital content producer Zach Silberberg, who has made waves on Twitter with his AI Biden speeches (one in which he discusses the Matt Damon film We Bought a Zoo, and another in which he complains of being “inverted” like the characters in Tenet), values anarchy highly.
“This version of the president is unpredictable,” Silberberg says. “We have no idea what the hell he’s going to say.” It could be silly, vulgar, rambling, or anything at all. He’s not slow and boring and more of the same neoconservative version of a Democrat we’ve been conditioned to accept as the most liberal version of a president we can get.”
Ultimately, Silberberg explains, the Biden riffs are about getting one over on the most powerful man in the world. Silberberg has also used ElevenLabs’ software to create humorous fake dialogues between Joe Rogan and Ben Shapiro. If we had a president like that, we could all laugh at him and torture him at the same time. Who wouldn’t want to watch Joe Biden get trapped in a haunted house? In a nutshell, it’s hilarious. Silberberg elaborated, saying, “AI is bad” and that he is “anti-AI.” His videos can be seen as cautionary tales in this regard.
Even though ElevenLabs is proud of their less impressive technological achievements, like making a fully synthetic voice that can read The Great Gatsby out loud, a large number of their users are more interested in pranks, whether they are funny or mean. The startup has thought about approving or rejecting each voice cloning request by hand, but if they don’t do this, there may not be much they can do to stop voice clones of our past and future presidents. That could be bad for democracy, but on the bright side, it means we get to listen to Trump and Biden argue like teenagers playing a first-person shooter online. Which, perhaps, describes our current political climate more accurately than we’d like to admit.