Staff Reports
The article alleges that tagging artificial intelligence-generated images with invisible markers might not be as foolproof as the government intends. This makes sense to this conservative writer because, let’s be real, when have government solutions ever worked flawlessly? The White House and AI developers are all gung-ho about using watermarking technology to combat fake images and misinformation. According to the Washington Examiner, Facebook and OpenAI are boasting about adding watermarks to photos to help sniff out those pesky “deepfake” images. But hold up, folks – according to experts in the field, this whole watermarking business might not be the silver bullet they’re hoping for.
Soheil Feizi, a computer science professor at the University of Maryland, argues that watermarks can easily be erased from AI-generated content, rendering them unreliable. Major players in the AI industry are teaming up to establish common watermarking standards, but these efforts might be futile. Anne Neuberger, the deputy national security adviser for cyber and emerging technology at the White House, is hyped about watermarking, but the skeptics are not having it.
Don't count on watermarking to prevent artificial intelligence deepfake election chaos pic.twitter.com/n3ELqvPR4q
— NY (@linyiq) February 12, 2024
This Feizi character insists that adversaries like China or Iran could effortlessly remove watermarks from AI-generated images, or even tamper with real images to appear watermarked. It seems like this whole watermarking thing might not be the brilliant solution it’s cracked up to be. Even the CEO of voice verification service Pindrop, Vijay Balasubramaniyan, has his doubts, pointing out that watermarks can fade away as images or audio are copied and transferred.
But fear not, dear readers. There may still be hope on the horizon! Feizi suggests linking social media images to their sources, allowing users to suss out any potential foul play. Perhaps in the future, researchers will come up with unstrippable watermarks, but for now, it looks like we’re stuck with this flawed technology. Welcome to the world of AI, folks – where the only thing certain is uncertainty.
