At a hearing In June 2019, experts from the House Intelligence Committee warned of the democracy-distorting potential of videos generated by artificial intelligence, known as deepfakes. Chairman Adam Schiff (D-California) played a clip in which Senator Elizabeth Warren (D-Massachusetts) was faked and urged social media companies to take the threat seriously because “after viral deepfakes the 2020 elections polluted, it will be too late by then. “Danielle Citron, then a law professor at the University of Maryland, said,” Fake video and audio could undermine the democratic process by influencing an election. “
The 2020 campaign is now history. There was upset, but deepfakes didn’t help. “Not really, no,” says Giorgio Patrini, founder of the deepfake tracking startup Sensity. Angie Hayden, product director at the AI Foundation testing a deepfake detection tool with media organizations and nonprofits like the BBC, also reported on a quiet campaign. “It’s nice if your technology saves the day, but it’s better if the day doesn’t have to be saved,” she says.
Lots of disinformation swirled around and still swirled around the recent vote, but the misleading videos that contributed to it appeared to be artisanal, not algorithmic. Fact checkers found videos deceptively described or edited with traditional tools, like a clip edited to make it look like Joe Biden had greeted Floridians as Minnesotans. An AI-generated profile photo attached to a fake person taking a disheveled and discredited smear against Biden’s son was revealed, but it only played a peripheral role in the stunt.
Twitter and Facebook added specific rules for deepfakes to their moderation guidelines in early 2020, but neither appear to have applied them. A Twitter blog post that rounded off its election efforts last week said it had labeled 300,000 tweets warning of misleading content since Oct. 27, accounting for 0.2 percent of all election-related posts over that period. There was no mention of deepfakes, and a company spokesman said he had “nothing specific” on the subject. Facebook did not respond to a request for a comment.
Two fake video campaigns attempted to convince US voters did so openly to warn of the technology’s potential.
Phil Ehr, a Democratic House candidate in the Florida Panhandle, posted a campaign ad featuring a fake version of incumbent Republican Matt Gaetz, in which he said unusual phrases like “Fox News sucks” and “Obama is way cooler than me” . Ehr’s own face – apparently completely human – breaks in to deliver a PSA about deepfakes and nation-state-backed disinformation. “If our campaign can produce a video like this, just imagine what Putin is doing,” he says.
The WIRED guide to artificial intelligence
Supersmart algorithms don’t do all of the work, but they learn faster than ever and do everything from medical diagnostics to showing ads.
Campaign advisor Keith Presley says Ehr, a Navy veteran who worked on electronic warfare, wanted Gaetz to look into the issue of disinformation, which Ehr believed Gaetz downplayed. The campaign contacted RosebudAI, a startup that uses deepfake technology to create images and videos for fashion shoots and online retail. According to Presley, the campaign designed the 60-second commercial to minimize the chance of malicious reuse. The algorithmic Gaetz is only shown on a TV in voters’ living rooms, not full screen, and there are giveaway glitches. Gaetz’s office did not respond to a request for comment.
Despite its sophisticated advertising, Ehr lost hard. Presley says while no malicious deepfakes occurred during the campaign, educating people is still important. He pointed out a paradox of looking for the fruits of technology that are claimed to seamlessly mimic reality: “How would anyone know?”