Spectacular footage of Bomb Cyclone 2023 hitting the northeast coast of the United States, courtesy of … [+]
(c) Brandon Clement/WX Chasing, used with permission
Brandon Clement isn't afraid of a bit of headwind. The Emmy Award-winning videographer has been releasing some of the most compelling footage of extreme weather events from around the world for over a decade. His YouTube channel WX Chasing has over 77,000 subscribers, his videos have been viewed more than 100 million times, and his work is often featured in mainstream news reports and features.
But there is a system that Clement has observed that is more than just troubling him: a perfect storm of greed, technology and indifference that threatens his livelihood and that of almost every creator hoping to get their work published on platforms like YouTube, Instagram, X to monetize and Facebook.
“It's destroying my business, it's causing so much stress and anxiety in my head that I can't sleep and I can't stop thinking about it,” Clement said in a phone interview earlier this month.
The scourge causing trouble for the tornado chaser? Shadowy operations that pirate copyrighted footage and repackage it into clickbait on social media platforms operate under hundreds of cut accounts in dozens of languages, harnessing the power of generative AI to a degree that overwhelms human-generated content threatens.
Emmy Award-winning videographer Brandon Clement
Courtesy of Brandon Clement
“I stole certain video excerpts from more than 60,000 pages on Facebook,” he said. “Some of these pages have millions of subscribers and followers, others have none at all. But if you start dividing your views by 60,000, you simply can't make money. You can’t grow your audience and your content will be ruined by overexposure.”
Science YouTuber Kyle Hill shares Clement's concerns about the threat. While he has so far benefited from his audience's ability to “tell the difference between quality content and auto-generated blah,” he is extremely concerned that the bad guys are gaining ground and that the problem goes far beyond niche news, Science and documentary content. Yesterday's situation, in which X (formerly Twitter) was flooded with fake images of Taylor Swift, was a more spectacular and horrific example of the same toxic combination of generative AI tools in the hands of unscrupulous operators that are expanding the scale and reach of social platforms hijack for their own benefit.
“The crux of the problem is that these fraudsters can quickly generate and steal content,” he explained. “YouTube [and other platforms] grants them advertising revenue until the owner or another party claims the stolen content. And by getting literally dozens of channels to upload new videos every few hours, these actors can consistently make enough money to continue their operations before a single YouTuber (like me) has the time to track it down and claim it.”
Hill and Clement cited specific examples of YouTube channels that traffic in AI-generated fake science content in several videos describing the problem, including this one.
Copyright infringement has been a problem on content platforms since the first platform launched, but the emergence of generative AI for text, voice, images and video has vastly increased the ability of thieves to distribute hundreds or thousands of videos under AI. generated headlines and thumbnails designed to generate views and which often contain false, misleading or completely incoherent content automatically generated from Wikipedia posts or random web scraping.
Science YouTuber Kyle Hill, 2024
Courtesy of Kyle Hill
“In the short term, I'm worried about just being drowned out by the nonsense,” Hill said. “There isn't enough time in the day to separate the good from the bad, especially if you just want to quickly see something during your lunch break. It's an old disinformation tactic. You do not have to lie; You just have to pollute the well so badly that no one cares about it anymore.”
Jevin West is an associate professor at the University of Washington Information School and co-founder of the Center for an Informed Public, which studies how false and harmful ideas spread and are amplified in the digital universe. “The real question is: Do users want a real person behind this content? And the data is inconclusive on that,” he said, noting that current data does not yet indicate an increase in the spread of misinformation since generative AI tools became mainstream in 2022-2023. “The danger is that when there is an information vacuum, such as natural disasters or elections, opportunists intervene. I expect it to get worse.”
Not all AI-generated content is stolen, fake, or problematic. In some cases, human creators use the tools to improve their own content or achieve greater production value. “If these videos used AI tools to create something never before done, something that enhanced human creativity or enabled someone like me to do something groundbreaking, then they should be allowed to compete,” explained Hill. “But that's not what it is all about here [pirate] Videos are. They are text-to-speech Wikipedia entries about stolen footage from everywhere from NetflixNFLX to the Weather Channel to you.”
In response to these issues, YouTube recently updated its terms on responsible AI innovation by making users more aware of content that contains AI-generated elements. “Specifically, we will require creators to disclose when they have created altered or synthetic content that is realistic, including through the use of AI tools,” wrote YouTube executives Jennifer Flannery O'Connor and Emily Moxley in a blog post dated November 2023. “When creators upload content, they will have new options to indicate that it is realistically altered or synthetic material. This could be, for example, an AI-generated video that realistically depicts an event that never happened, or content that shows someone saying or doing something they never actually did.”
The new policy also includes a provision that requires the removal of content that does not meet standards of decency or violates the privacy of individuals by displaying it without permission. However, the post states: “Not all content will be removed from YouTube and we will consider a number of factors when evaluating these requests.”
For YouTubers who are victims of systematic hijacking of their content, like Brandon Clement, that's not good enough. “Most of these platforms have tools that enable content discovery and give creators the right to act on it, but they only allow access to major production houses and major music labels,” he said. Meanwhile, for ordinary YouTubers, filing takedown requests against bad actors is so time-consuming, cumbersome, and often fruitless that perpetrators can make most of their money by removing their videos or making them private before they show up on the platform's radar. As we saw yesterday with the Taylor Swift fakes, bad content can spread quickly while countermeasures take time, even when the victim is one of the world's biggest and most influential celebrities.
“They are allowed to delete evidence while the investigation is ongoing,” he said. “There is no penalty. They are rewarded because they can avoid punishment. It would be very easy for YouTube to prevent any action on a video once a DCMA is in place [Digital Millennium Copyright Act] Application is submitted. If they wanted to stop it, they could.”
Clement also noted that the platforms benefit from engagement and clicks regardless of where they come from, and may have incentives to allow more AI-generated, algorithmically optimized, synthetic content if it outperforms metrics of human-generated work. YouTube and its parent company GoogleGOOG declined to make a spokesperson available to address these concerns.
Facing the platforms' indifference, Clement has taken matters into his own hands, founding a company called ViralDRM to advocate for creators in legal and procedural actions, and filing DMCA action against perpetrators, including recently against Indian news channel News Nation, TV9 Bharatvarsh and Zee News.
Despite these efforts, the legal and regulatory systems are blunt instruments against such nimble and fast-moving technology, especially when there is no international consensus on how to deal with the problem.
“This is not an easy solution even for a company that has more cash than some countries,” West said. “What they can do more than anything is advance efforts, for example by watermarking anything that is created synthetically. This may only represent 80% of the content, so there will be a significant portion that does not adhere to these standards.”
Kyle Hill believes there may be some alternatives for creators and conscious consumers. “Most of the bigger YouTubers I know have ways to support them directly and that can be a big help. Patreon, YouTube memberships, merchandise, etc. But while there are some benefits to alternative media and revenue streams outside of the ad-supported model, I still fear that the fragmentation of our information ecosystem will do more harm than good. More spam, more lies, more mistrust, more extremism, more division.”
Comments are closed.