In the age of generative artificial intelligence, not everything can be taken at face value. As AI video apps become more sophisticated, the internet is full of ultra-realistic AI videos that are indistinguishable from reality. Fortunately, AI experts say there are several ways to tell if what you're seeing is real or a very convincing fake.
Check watermark
Listen to garbled audio
AI experts say there are “clear signs” that “voices and sounds in AI videos often reveal their synthetic origins.” HuffPost. While some words are spoken more slowly than others due to the natural rhythms of real-life conversation, “AI voices often always sound unnaturally rushed.” As people have devised ways to find AI-generated content, the em dash has become synonymous with ChatGPT-generated text. When asked about the equivalent in the video, Sora director Bill Peebles said, “Sora likes to say a lot of words quickly, this slightly wired speech pattern.” interview Includes video streaming program TBPN.
this week
The Week provides readers with a wide range of perspectives from 200 trusted news sources.
Try 6 free questions
Sign up for this week's free newsletter
From our daily WeekDay news briefing to our award-winning food & drink emails, you'll get the best of the week delivered straight to your inbox.
From our daily WeekDay news briefing to our award-winning food & drink emails, you'll get the best of the week delivered straight to your inbox.
According to HuffPost, AI-generated voices have not yet learned the speaking rhythms of natural voices, so app-generated voices often produce “garbled sounds that appear to flatten the pitch of natural sounds.” Melissa Bass-Burke, a linguistics professor at the University of Chicago, told HuffPost that humans “will never produce the same kind of garbled quality, because we literally can't do it.” Our vocal tracks cannot “move from one note to another” without “blurring the information between the two notes”.
While it may seem tedious, checking a video's metadata reveals its origin and is “easier than you might think,” he said. CNET. Metadata is automatically associated with content upon creation and may include “the type of camera used to take the photo, location, date and time the video was captured, and file name.” All photos and videos online, “whether created by humans or AI,” contain metadata. Many AI-generated videos also include “content credentials indicating the origin of the AI.”
Consider content validity and sources
One of the easiest ways to detect slop in AI is to ask whether what you're seeing is really possible, said Zhuang Liu, a computer science professor at Princeton University. rolling stone. If something “couldn't happen in the real world”, it was clearly generated by an AI. For example, “a horse on the moon or a chair made of avocado.” Impossible means “these are clearly generated by AI,” he says. “That's the simplest case.”
Next, check the source where you found the image. This “doesn't necessarily work well for viral content,” especially since it “often comes from previously unknown accounts,” but “seeing a video on a meme page can be a clue that it's not real,” Rolling Stone said.
Please continue to be cautious
Unfortunately, “there is no foolproof way to accurately tell at a glance whether a video is real or AI,” says CNET. The best way to avoid being fooled is to not automatically and unquestionably believe everything you see online. Trust your intuition. If an item “feels unreal,” it probably is. In this “unprecedented era of AI disruption,” the best thing to do is to “take a closer look at the videos you're watching.”
See more details
