Up close and personal with AI 'fake news'

How the next wave of technology is upending the global economy and its power structures

Apr 16, 2024 View in browser   By Derek Robertson Fabrice Coffrini/AFP via Getty Images Generative AI will have a massive impact on global politics, already sowing discord with fake news, fake debates, and even fake avatars for real candidates. But how bad could it really get? In a reporting series published today, POLITICO’s Mark Scott finds some cause for alarm. The U.S. presidential race is just one of 50 elections happening this year, and AI-generated images and audio meant to mislead, persuade or just plain confuse voters are poised to sweep across the globe as governments struggle to catch up. Mark invokes a litany of such examples, including but not limited to: Fake audio of British Labour Party leader Keir Starmer berating his staff; Robocalls with an AI-generated “Joe Biden” discouraging voters from participating in a New Hampshire primary; a fake “Tom Cruise” criticizing the Paris Olympics as part of a Russian propaganda effort; and even deepfake pornography of a Northern Ireland assembly member. And the companies helping to develop powerful AI models are struggling to keep up: “People are constantly coming up with new ways to try [to] break the systems,” Natasha Crampton, chief responsible AI officer at Microsoft, told Mark. (Microsoft is a major backer of both OpenAI and Mistral, putting it at the center of efforts to combat false or harmful AI-generated content.) But Mark also found in interviews with national security officials, tech executives and watchdog groups that while they view AI-generated deepfakes as a vexing tool for bad political actors, it’s also just that — one tool in the toolbox. Josh Lawson of the Aspen Institute’s AI Elections Initiative told Mark his group has found little concrete evidence of AI fakes directly changing voter behavior, and another researcher said that more egregiously fake content like imagery of former President Donald Trump surrounded by gleeful Black supporters wasn’t “successful in the wild.” Mark and POLITICO visual producer Lucia Mackenzie captured the public perception of AI-generated election material in a series of visual charts based on Ipsos polling, finding significant global divides when it comes to how people view the technology. Notably, they find that people in developing economies where the disruption of the smartphone is still a recent memory say they understand AI better than their counterparts in developed countries, and are more sanguine about AI overall to boot. Those people are also more confident in their ability to distinguish between real news and AI deepfakes. The top five countries for confidence in that ability (in descending order) are Indonesia, South Africa, Turkey, Thailand, and Brazil; the bottom five are France, Spain, the United States, Germany, and South Korea. Mark and Lucia also drew on polling by KFF, and the Associated Press and NORC Center for Public Affairs Research about trust in online media specifically in the U.S., and the results were not encouraging. Nearly 31 percent of U.S. adults, according to the KFF poll, “report feeling uncertain about the accuracy of information about the upcoming 2024 presidential election all or most of the time.” A majority of adults surveyed thought AI would bolster the spread of false information, with those more knowledgeable about AI more likely to believe this was true. …But what about you, dear reader? Forget the polls: Another part of the series published today puts your own ability to distinguish between real and AI-generated content to the test, comparing real images of world leaders and events with a series of fake ones generated by Mark using Midjourney. (Full disclosure: As someone who is, to put it reductively, paid to look at AI-generated images for much of my day, I only got 44 percent correct.) There are also a pair of sound clips comparing Mark and POLITICO editor Aoife White’s voices with AI-generated clones, although for my money, those are less convincing. When it comes to video, the team put a pair of questions to the infamous, ongoing fake livestreamed “debate” between Biden and Trump, resulting in a stream of the irreverent and vulgar retorts for which that channel is known. Those answers are purposely fake. But hazy, general estimations of a politician’s character often play more of a role in opinion formation than their actual character. For AI deepfakes to make a difference in such an environment they might not need, necessarily, to change voters’ minds — but only to help them confirm what they already wanted to believe.   THE GOLD STANDARD OF POLICY REPORTING & INTELLIGENCE: POLITICO has more than 500 journalists delivering unrivaled reporting and illuminating the policy and regulatory landscape for those who need to know what’s next. Throughout the election and the legislative and regulatory pushes that will follow, POLITICO Pro is indispensable to those who need to make informed decisions fast. The Pro platform dives deeper into critical and quickly evolving sectors and industries—finance, defense, technology, healthcare, energy—equipping policymakers and those who shape legislation and regulation with essential news and intelligence from the world’s best politics and policy journalists. Our newsroom is deeper, more experienced, and better sourced than any other—with teams embedded in the world’s most active legislative and regulatory power centers. From Brussels to Washington, New York to London, Sacramento to Paris, we bring subscribers inside the conversations that determine policy outcomes and the future of industries, providing insight that cannot be found anywhere else. Get the premier news and policy intelligence service, SUBSCRIBE TO POLITICO PRO TODAY.     more vr in the classroom Meta is continuing its push for virtual reality in classrooms, announcing a new VR product that will be solely designed for educational use. The company announced yesterday that the platform, which it compared to its software suite for business in VR (but has no name or release date as of yet), will “allow teachers, trainers and administrators to access a range of education-specific apps and features, and make it possible for them to manage multiple Quest devices at once,” bypassing some of the tech’s logistical hurdles. The platform will be available for users aged 13 and older. Announcing the new product, Meta’s president for global affairs Nick Clegg pointed to recent research from Morehouse College that showed VR tools boosting academic performance. doom (and boom) AI “doomers” might fret about the next powerful LLM ending humanity, but worrying about new tech’s eschatological implications is a surprisingly old pastime. One relatively recent example was a circa-2000 essay in Wired from Sun Microsystems co-founder Bill Joy titled “Why the Future Doesn’t Need Us,” which was exhumed by American Enterprise Institute fellow James Pethokoukis in a recent blog post. Joy argued in his essay — at a time when many people still would have been reading it with the assistance of America Online — that the “perfection of extreme evil” wrought by tech would lead to a “surprising and terrible empowerment of extreme individuals.” Pethokoukis, a hardcore techno-optimist, argues that this foreshadowed how “dystopian images of the tomorrow undercut efforts at tech progress in the present,” something he argues is reflected in current regulation efforts “that focus on the risks and limitations of innovation.” As Pethokoukis also points out, viewpoints tend to diverge wildly at moments of major technological change — like both the Y2K-era boom in internet access and today’s AI revolution. Not long before Joy’s essay Wired co-founder Kevin Kelly proclaimed to the reader that “The good news is, you’ll be a millionaire soon” — a promise that should sound familiar to optimistic early adopters of another disruptive new technology. Tweet of the Day The Future in 5 links Microsoft is making a $1.5 billion investment in the UAE that might carry some geopolitical risk. Chinese tech titan Baidu claims its chatbot now has 200 million users. Meta’s oversight board is investigating its handling of AI-generated deepfake nudes. Investors might already be tiring of AI. Learn the secret history of WWII-era scientists experimenting on themselves. Stay in touch with the whole team: Derek Robertson (drobertson@politico.com); Mohar Chatterjee (mchatterjee@politico.com); Steve Heuser (sheuser@politico.com); Nate Robson (nrobson@politico.com); Daniella Cheslow (dcheslow@politico.com); and Christine Mui (cmui@politico.com). If you’ve had this newsletter forwarded to you, you can sign up and read our mission statement at the links provided.   POLITICO IS BACK AT THE 2024 MILKEN INSTITUTE GLOBAL CONFERENCE: POLITICO will again be your eyes and ears at the 27th Annual Milken Institute Global Conference in Los Angeles from May 5-8 with exclusive, daily, reporting in our Global Playbook newsletter. Suzanne Lynch will be on the ground covering the biggest moments, behind-the-scenes buzz and on-stage insights from global leaders in health, finance, tech, philanthropy and beyond. Get a front-row seat to where the most interesting minds and top global leaders confront the world’s most pressing and complex challenges — subscribe today.       Follow us on Twitter Ben Schreckinger @SchreckReports Derek Robertson @afternoondelete Steve Heuser @sfheuser   Follow us

To change your alert settings, please log in at https://login.politico.com/?redirect=https%3A%2F%2Fwww.politico.com/settings This email was sent to rouf@idiot.cloudns.cc by: POLITICO, LLC 1000 Wilson Blvd. Arlington, VA, 22209, USA Unsubscribe | Privacy Policy | Terms of Service