AI-Generated videos Are Flooding the Internet-Here’s How to Spot the Fakes
A surge in refined artificial intelligence video generation is creating a flood of “AI slop”-low-quality, frequently enough deceptive content-making it increasingly difficult to distinguish reality from fabrication. Experts warn viewers to be vigilant, looking for telltale signs of AI involvement as the technology rapidly evolves.
Several key indicators can reveal an AI-generated video. Inconsistent lighting, unrealistic facial expressions (such as overly large smiles or strangely open eyes while crying), and unnaturally smooth or “airbrushed” skin are common flaws. Perhaps the most noticeable issue is the tendency for background details to disappear or morph throughout the video. A recent example cited involved Taylor Swift‘s promotional material for her new album,where the spokes of a Ferris wheel in the background appeared blurred as it rotated.
Currently, OpenAI attempts to mitigate deception by watermarking videos created with its Sora platform, though pro users can download versions without the watermark.Tutorials demonstrating how to remove these watermarks are already circulating on platforms like YouTube.
While companies like OpenAI and Meta claim to prioritize transparency, the potential for misuse remains important. “It’s hard because by the very nature of technology like this, it’s going to be misused,” one expert noted. ”So you just have to see if you can stem that misuse as much as possible…But we’re going to have to wait and see how triumphant they are at that. And right now, if history is any guide, I’m a little concerned.”