Taipei, Taiwan – “The American Dream. They say it's for everyone, but is that really true?”
So begins a 65-second, AI-generated animated video that touches on hot-button issues in the United States, from drug addiction and incarceration rates to the widening gap between rich and poor.
Storm clouds gather over a cityscape resembling New York City, and as the video ends, the words “AMERICAN DREAM” appear in the darkening sky.
The message is clear. Despite the promise of a better life for all, the United States is in terminal decline.
The video, titled “American Dream” or “American Mirage,” was aired by Chinese state broadcaster CGTN as part of the animated series “A Fractured America,” and features several segments that were widely shared on social media. One of the.
Other videos in the series include “American Workers in Trouble: Unbalanced Political and Economic Consequences” and “Revealing the Real Threat: America's Military-Industrial Complex” Contains similar titles that evoke images.
In addition to loud anti-American messages, all the videos share a highly stylized AI-generated aesthetic and eerie computer-generated audio.
CGTN and the Chinese embassy in Washington, D.C., did not respond to requests for comment.
American workers in turmoil: Unbalanced political and economic outcomes #firstvoice pic.twitter.com/JMYTyN8P2O
— CGTN (@CGTNOfficial) March 17, 2024
The Fractured America series explores how artificial intelligence (AI), which can generate high-quality multimedia in seconds with minimal effort, is beginning to shape Beijing's propaganda efforts to undermine the United States' standing in the world. This is just an example.
Henry Ajdar, a UK-based generative AI expert, says the CGTN series doesn't pretend to be real video, but it does show how AI has made it easier and cheaper to mass-produce content. He said that this is a clear example of this.
“The reason they did it this way is because they could have hired animators and voice-over artists to do this, but it would probably take more time in the end. It will cost more,” Ajder told Al Jazeera.
“This is a cheaper way to scale up your content creation. If you can combine all of these different modules, you can generate images, animate those images, or just generate videos from scratch. You can generate very engaging and human-like speech synthesis, which means you have an entire automated, or at least highly synthetically generated, content creation pipeline.”
China has long used the internet's vast reach and borderless nature to conduct influence campaigns abroad.
China's vast army of internet trolls, known as “wumao,” became known more than a decade ago for flooding websites with stories about the Chinese Communist Party.
Since the advent of social media, the Chinese government's propaganda efforts have come to rely on platforms such as X and Facebook and online influencers.
As Black Lives Matter protests swept across the United States in 2020 following the killing of George Floyd, the Chinese government restricted criticism of its record of discrimination against ethnic minorities such as Uyghur Muslims in the country. However, Chinese state social media accounts expressed support.
“I can not breathe.” pic.twitter.com/UXHgXMT0lk
— Hua Chunying 华春莹 (@SpokespersonCHN) May 30, 2020
Microsoft's Threat Analysis Center said in a report last year that the introduction of AI has made it easier to create viral content and, in some cases, more difficult to identify that the content was created by a nation-state actor. Ta.
According to Microsoft, the Chinese government-backed organization has been deploying AI-generated content since at least March 2023, and such “relatively high-quality visual content is already receiving higher rates from genuine social media users.” “It brings out a high level of engagement.”
“Over the past year, China has been automatically posting images that can be used for influence operations aimed at mimicking U.S. voters across the political spectrum and provoking controversy along racial, economic, and ideological lines. “We have honed new capabilities to generate new capabilities,” the report said.
“This new feature is powered by artificial intelligence that seeks to create high-quality content that can be spread across social networks in the United States and other democracies.”
Microsoft also identified more than 230 state media employees posing as social media influencers with a reach of 103 million people in at least 40 languages.
Their stories followed a similar script to the CGTN video series. China is rising and winning the race for economic and technological superiority, while the United States is collapsing and losing friends and allies.
As AI models like OpenAI's Sora generate increasingly surreal videos, images, and audio, identifying AI-generated content will become increasingly difficult, fueling the proliferation of deepfakes. right.
Astroturfing, the practice of building broad social consensus on a particular issue, is poised for “revolutionary improvement,” according to a report published last year by Rand, a think tank partially funded by the U.S. government. There is a possibility that it will be prepared.
The CGTN video series, while using sometimes awkward grammar, captures many complaints shared by U.S. citizens on platforms like X, Facebook, TikTok, Instagram, and Reddit (websites where AI models collect data for training). is reflected.
Microsoft said in its report that while the advent of AI does not make it any more likely that the Chinese government will interfere in the 2024 U.S. presidential election, it does make it more likely that “if the Chinese government decides to intervene, it will make election interference more likely.” “It's very likely to be effective.” .
The United States is not alone in worrying about the potential for AI-generated content and space lawns as we head into a turbulent election year.
By the end of 2024, elections affecting 2 billion voters will be held in more than 60 countries in a record year for democracy.
This includes democratic Taiwan, which elected new president William Lai Ching-toku on January 13th.
Taiwan, like the United States, has a questionable political status and is a frequent target of Beijing's influence operations.
Beijing claims Taiwan and its remote islands as part of its territory, even though it functions as a de facto independent state.
In the run-up to the January general election, more than 100 deepfake videos of fake newscasters attacking outgoing Taiwanese president Tsai Ing-wen were created by China's Ministry of State Security, the Taipei Times reported, citing national security officials. It was reported as.
Like the CGTN video series, these videos lacked sophistication, but they showed how AI can help spread misinformation on a large scale, said co-director of the Taiwan Information Environment Research Center (IORG). , said Chihhao Yu.
Yu said his organization tracked the spread of AI-generated content on LINE, Facebook, TikTok and YouTube during the election period and found that AI-generated audio content was particularly popular.
“[The clips] They are often spread through social media and framed as leaked/secret recordings of politicians or candidates regarding personal affairs or corruption scandals,” Yu told Al Jazeera.
AI expert Ajder said deepfake audio is harder for humans to distinguish from the real thing compared to doctored or AI-generated images.
A recent incident in the UK, where a general election is scheduled for the second half of 2024, featured opposition leader Keir Starmer appearing in a deepfake audio clip of him abusing his staff.
Such convincing misrepresentation would not have been possible before without “impeccable impressionism,” Ajder said.
“Actors aligned with or connected to the state that have a motive or a potential goal to achieve now have new tools to try to achieve it.” Ajdel said.
“And some of those tools can help us extend what we're already doing. But in some situations, we can do those things using entirely new means that are already difficult for governments to address. It might help you achieve that.”