With 10 months until Election Day, artificial intelligence is poised to change the nature of American politics — and the tech is so easy to use, even a reporter can do it.
A recent robocall to New Hampshire residents featured a fake Joe Biden voice urging Democrats to skip Tuesday’s primary. He didn’t really say that, and the New Hampshire attorney general’s office launched an investigation into the call, with a spokesperson saying it appeared to be an unlawful attempt to disrupt the primary and “suppress New Hampshire voters.”
Artificial intelligence has advanced quickly in recent years, raising fears of efficient, widespread misinformation campaigns, to the detriment of voters and news consumers.
The uses for the technology in politics are seemingly endless, from astroturfed comment sections and legislative correspondence to entirely AI-generated podcasts, which the firm NewsGuard found plagiarized from real news websites. In the future, fake information like the Biden robocall could be produced on a mass scale — and, crucially, employed in local elections or in languages other than English, making the truth even more difficult to ferret out. Imagine receiving a call that your local polling place is closed because of a busted water line.
So HuffPost tried our hand at the deceptive practice.
It only took a few minutes, and the results are strikingly realistic.
Here, for example, is a fake “Joe Biden” announcing that cannabis is now legal nationwide:
And here’s our AI-powered “Donald Trump” reciting Julia Stiles’ character Kat Stratford’s iconic poem from “10 Things I Hate About You”:
The spoofs were created with Parrot AI, a popular app that offers users the opportunity to put words in the mouths of dozens of prominent politicians, celebrities, and even cartoon and video game characters. The company did not return HuffPost’s request for comment about what rules were in place, if any, to prevent the app from being used to mislead voters.
Given the availability of the technology, many observers are concerned about the so-called “liar’s dividend,” or the idea that politicians can falsely dismiss inconvenient facts as fake AI material. The preponderance of artificial material could also degrade Americans’ already-fragile sense of a shared reality — a problem that’s been discussed by top experts in the field.
“In a democratic society, if people just stop believing anything, then it’s eroding really a core tenet of a democratic system, which is trust,” Sarah Kreps, a political scientist who’s studied artificial intelligence, said during a meeting of the President’s Council of Advisors on Science and Technology last year.
In that context, there’s a crucial difference between being aware of misinformation — and feeling debilitated by it.
“Voters should know these tools exist, but also that the facts still matter and that the truth is knowable,” said Josh Lawson, director of AI and Democracy at the Aspen Institute. “That is extremely important. We know that bad actors want to erode our confidence in truth, and the knowability of facts.”
Rather than looking for some “telltale sign” that a piece of content is fake — that’s a tall order, especially with the technology getting more sophisticated and widespread — Lawson offered some straightforward advice: “When there’s information that calls for a reaction from you, verify that with trusted information sources” like reliable news media or public officials, especially if the information in question could change how or whether you vote.
Other actors in the AI field have used absurdism to make a similar point: On Twitch, the livestreaming video site, artificial intelligence bots of Donald Trump and Joe Biden have for months now been engaged in a continuous, digitally improvised “debate” in which they insult each other and respond in real time to comments from a chatroom watching the spectacle. (“No amount of tanning can hide the darkness in your soul,” Biden told Trump Wednesday afternoon, one in an endless string of snapbacks.)
TrumpOrBiden2024, the channel in question, is a project of MySentient.AI, which also runs an AI “Ask Jesus” stream. Reese Leysen, the company’s CEO, told HuffPost that one initial goal of the stream was showcasing just how realistic the technology has become.
“While we are very clearly portraying it as parody — labeling it as such, making it so wacky and unhinged that obviously it’s parody — the effect that you have… is so many people find it and say, ‘Wait a minute, this is pretty damn convincing.’ It does make people more aware of the fact that things they bump into like that may not be real.”
Leysen compared the current era of artificial intelligence to the introduction of Photoshop, when the public eventually realized that a given image may have been digitally manipulated. Still, he said, “at the moment, it’s already more omnipresent than a lot of people may be aware of.” AI material, he noted, is clogging search engines and X, formerly known as Twitter.
“It’s very quickly spreading,” Leysen said. “It’s going to be interesting to see how we adapt.”
