Deepfakes Barely Impacted 2024 Elections Because They Aren’t Very Good, Research Finds

It seems the internet is growing though Drowning in fake imagesWe can at least take some stock in humanity’s ability to smell BS when it matters. A number of recent studies suggest that AI-generated disinformation has not had a material impact on this year’s elections around the world because it is still not very good.

There has been much concern over the past year that increasingly realistic but synthetic content can manipulate viewers in harmful ways. The rise of generative AI has reignited those fears, as the technology makes it easier for anyone to create fake visual and audio media that seems real. In August, a political consultant used AI President Biden’s Voice Fraud A robocall for New Hampshire voters to stay home during the state’s Democratic primary.

Tools like ElevenLabs make it possible to submit a short soundbite of someone speaking and then mimic their voice to say whatever the user wants. Although many commercial AI tools include guardrails to prevent this use, open-source models are available.

Despite this progress, Financial Times A new story looking back on the year found that very little synthetic political content went viral around the world.

This is a quote Report The Alan Turing Institute found that only 27 pieces of AI-generated content went viral during the summer European elections. The report concluded that there was no evidence that the elections were influenced by AI disinformation because “most exposure was concentrated among a minority of users whose political beliefs were already aligned with ideological narratives embedded in such content.” In other words, among the few who saw the content (perhaps before flagging) and were primed to believe it, it reinforced those beliefs about a candidate even as those familiar with it knew the content itself was AI-generated. It cited an example of AI-generated footage showing Kamala Harris addressing a rally standing in front of a Soviet flag.

In the US, the News Literacy Project identified more than 1,000 examples of misinformation about the presidential election, but only 6% were done using AI. In X, references to “deepfake” or “AI-generated” in the community notes are usually only made with the release of a new image generation model, not during selection.

Interestingly, it appears that users on social media were more likely to misidentify real But in general, users displayed a healthy dose of skepticism when comparing images as AI-generated the other way around. And fake media can still be debunked through official communication channels or other means like Google reverse image-search.

If the results are correct, it will make a lot of sense. AI imagery is everywhere these days, but images created using artificial intelligence still have an off-putting quality to them, exhibiting tell-tale signs of being fake. An arm may be unusually long, or a face not properly reflected on a mirrored surface; There are many small clues that tell an image is synthetic. Photoshop can be used to create much more convincing fakes, but it takes skill to do so.

AI proponents shouldn’t cheer at this news. This means that the generated images still have a way to go Anyone checked out OpenAI’s Sora model knows that the video it produces isn’t very good – it almost looks like something made by a video game graphics engine (It is assumed that it was trained on video games), which clearly do not understand physics-like properties.

All that being said, there were still concerns. Alan Turing Institute Report did After all, it concludes that beliefs can be reinforced by a realistic deepfake that contains false information even though the audience knows the media isn’t real; Confusion about whether a piece of media is authentic damages trust in online sources; And AI imagery is already used Targeting female politicians with pornographic deepfakesWhich can be damaging psychologically and to their professional reputation as it reinforces sexist beliefs.

Technology will definitely continue to improve, so keep an eye on it.