In an era of deepfakes, bot-generated books, and AI images created in the style of famous artists, the promise of digital watermarks to identify AI-generated images and text has been tantalizing for the future of AI transparency.
Recently, companies and researchers have been exploring the potential of digital watermarks, including a new watermarking tool called SynthID by Google DeepMind. This tool embeds a digital watermark directly into the pixels of an image, making it imperceptible to the human eye but detectable for identification.
However, it’s important to note that digital watermarks, whether visible or invisible, are not sufficient to stop bad actors. As Soheil Feizi, a computer science professor at the University of Maryland, stated, “we don’t have any reliable watermarking at this point — we broke all of them.” Feizi and his fellow researchers have discovered various ways for bad actors to evade watermarking attempts.
