new york
CNN
—
Last month, a video posted on Twitter by Florida Gov. Ron DeSantis’ presidential campaign showed former President Donald Trump hugging Dr. Anthony Fauci, apparently generated by artificial intelligence. was used. The images, supposedly intended to criticize President Trump for not firing the country’s top infectious disease expert, were hard to find. These images were displayed alongside real-life images of the two, overlaid with the text “Trump in real life.”
As the image began to go viral, it was quickly flagged as fake by fact-checking agencies and discerning users. But Twitter, which has cut much of its staff in recent months under new ownership, did not remove the video. Instead, they ended up adding a community note (a contributor-driven feature that highlights misinformation on social media platforms) to the post, with a video that read, “Three pictures of President Trump hugging Fauci.” Still images are AI-generated images,” warned site users.
Digital information integrity experts say this is just the beginning of AI-generated content being used in ways that could confuse or mislead voters ahead of the 2024 US presidential election. claim.
New AI tools offer the ability to generate compelling text, realistic images, and even video and audio. Experts and even some executives overseeing AI companies say these tools risk spreading false information that misleads voters, including ahead of the 2024 US elections.
“The campaign is starting to kick in, elections are fast approaching, technology is advancing fast,” said Jevin West, a professor at the University of Washington and co-founder of the Center for Informants. . “We are already seeing evidence of the impact of AI.”
Experts say that as platforms accessed by billions of people for information and frequently used by bad actors to spread false claims, social media companies have a significant responsibility to address these risks. point out that it owes But they now face a perfect storm of factors that could make it harder than ever to respond to the wave of misinformation about the upcoming election.
Several major social networks have rolled back crackdowns on some election-related misinformation and have undergone significant job cuts over the past six months, potentially threatening election integrity, safety and accountability. A blow to an AI team. Current and former U.S. officials also said a federal judge’s decision earlier this month to limit the way some U.S. government agencies communicate with social media companies is a response to federal and state election misinformation. have warned that it could have a “chilling effect” on (On Friday, the Court of Appeals temporarily stayed the order.)
On the other hand, AI is evolving rapidly. And despite calls from industry insiders and others, U.S. lawmakers and regulators have yet to implement real guardrails against AI technology.
“We’re not even confident in our ability to deal with the older types of threats,” says David Evan Harris, an AI researcher and ethics adviser at the Institute of Psychological Technology. He previously worked on responsible AI at Facebook’s parent company, Meta. “And now there is a new threat.”
Major platforms have put in place existing policies and practices related to misinformation for CNN, in some cases specifically targeting “synthetic” or computer-generated content, and identifying AI-generated misinformation. He said it would help him deal with it. None of the companies agreed to interview anyone working on generative AI detection efforts.
Bhaskar Chakravorty, dean of global business at Tufts University’s Fletcher School, told CNN that the platform “wasn’t ready in the past, and there’s absolutely no reason to believe it will be ready now.” rice field.
Misleading content, especially election-related, is nothing new. But with the help of artificial intelligence, anyone can quickly, easily, and cheaply create masses of fake content.
And given the rapid advances in AI technology over the past year, by next year’s U.S. presidential election, it could become even more difficult to spot fake images, text, audio, and video.
“There is still over a year until the election. These tools will continue to improve and in the hands of sophisticated users they can be very powerful,” Harris said.. He added that AI will likely exacerbate the kinds of misinformation and election interference that happened on social media in 2016 and 2020.
The combined use of various forms of AI-generated content can make false information more believable. For example, AI-generated fake articles accompanied by AI-generated photos purporting to show what happened in the report, said researcher Margaret Mitchell. Chief ethics scientist at the open source AI company Hugging Face.
Andrew Angeloff/Alamy Stock Photo
Experts worry that the proliferation of generative AI tools like ChatGPT will make it easier for bad actors to create misinformation about elections.
AI tools are useful for those who want to mislead, but especially for organized groups and foreign adversaries motivated to intervene in US elections. Massive foreign troll farms have been hired to influence past elections in the United States and elsewhere, but now “one person can generate thousands of functioning troll farms” to pump content across social media. I’ll be in charge of deploying AI bots,” said Mitchell, who previously worked at Google, to mislead voters.
OpenAI, makers of the popular AI chatbot ChatGPT, issued a stark warning about the risks of AI-generated misinformation in a recent research paper. A large amount of misinformation from AI systems, whether intentional or generated by bias or “hallucinations” from the system, “casts the entire information environment into question and our ability to distinguish fact from fiction.” It could threaten our ability,” he said.
Examples of AI-generated misinformation are already beginning to emerge. In May, several Twitter accounts, including those who paid for the blue “authenticated” checkmark, shared fake images purporting to show an explosion near the Pentagon. The error in the image was quickly debunked, but the stock market fell briefly after the circulation. Twitter has suspended at least one of the accounts involved in spreading the image. Facebook has fact-checked and labeled the post about the image as “false information.”
A month earlier, in response to President Joe Biden’s official campaign announcement, the Republican National Committee released a 30-second ad that used AI imagery to imagine a dystopian United States after the 46th president’s re-election. announced. The RNC ad included a small disclaimer on the screen that read, “It’s built entirely using AI imagery,” but among potential voters in Washington, D.C., whom CNN showed the video to, Some people didn’t realize that on their first viewing.
Dozens of Democratic lawmakers last week sent a letter to the Federal Election Commission urging it to consider cracking down on the use of artificial intelligence technology in political advertising, saying deceptive ads could undermine the health of next year’s elections. warned that there is
Looking ahead to 2024, many platforms have announced plans to roll out plans to protect the integrity of elections from threats such as AI-generated content.
Earlier this year, TikTok introduced a Civil Integrity Policy, which prohibits misleading information about the election process, and a General Misinformation Policy, which prohibits false or misleading claims, along with an AI-generated “synthetic” policy. Or announced a policy stipulating that manipulated media must be clearly labeled. May cause “serious harm” to individuals or society.
YouTube has a manipulated media policy that prohibits content that is “manipulated or processed” in a way that misleads users and “may pose a significant risk of serious harm.” The platform contains content that may mislead users about when and how to vote, false claims that may discourage voting, and “encouraging others to interfere in the democratic process.” There is also a policy on content. YouTube also publishes credible news and information about the election prominently on its platform, and its election-focused teams include members of the Trust & Safety, Products, and “Intelligence Desk” teams. It also says that there are
“Technically manipulated content (such as election content) that could mislead users and cause serious harm is not allowed on YouTube,” YouTube spokeswoman Ivy Choi said in a statement. Stated. “We use machine learning and human review to enforce our policies on manipulated content, and we continuously improve this effort to stay ahead of potential threats.”
A Meta spokesperson told CNN that the company’s policy applies to all content on its platform, including AI-generated content. This includes removing false allegations that the platform may directly contribute to “interfering with the functioning of the political process and certain highly deceptive manipulated media” and reducing the spread of other misleading allegations. Includes a misinformation policy that states that it is possible. Meta also prohibits ads featuring content proven false by a network of third-party fact-checkers.
TikTok and Meta also joined a group of technology industry partners coordinated by the non-profit Partnership on AI, dedicated to developing frameworks for the responsible use of synthetic media.
When asked to comment on the article, Twitter automatically responded with a poop emoji.
Twitter has backed off much of its content moderation in the months since billionaire Elon Musk took over the platform, instead allowing users to criticize the accuracy of others’ posts and criticize other people’s posts. The focus is on the “community notes” feature that allows you to add context to your posts. Twitter has a “synthetic media” policy on its website that allows it to label or remove “synthetic, manipulated, or out of context media that may deceive, confuse, or harm people.” It also states that there is
Still, as is often the case with social media, the challenge will be less policy enforcement than policy enforcement. The platform primarily combines human and automated review to identify misinformation and manipulated media. Both companies declined to provide additional details about the AI detection process, including how many staff members are involved in such efforts.
But AI experts say they worry the platform’s detection system for computer-generated content will struggle to keep up with technological advances. Even some of the companies developing new generative AI tools struggle to build services that can accurately detect that something is AI-generated.
Some experts have urged all social platforms to introduce policies requiring clear labeling of AI-generated or manipulated content, and have called for regulators and lawmakers to impose guardrails around AI. and to hold tech companies accountable for the prevalence of false claims.
One thing is clear: the stakes for success are high. Experts say AI-generated content not only creates the risk of misinformation for internet users, it also makes it harder to trust real information about everything from polls to crisis situations. There is also the possibility of becoming.
“I know we’re in a very scary situation where it becomes very unclear what happened and what didn’t really happen,” Mitchell said. “When it comes to whether or not the content we are watching is authentic, it completely destroys the foundation of reality.”
