>Based on our discussion about the different types of image hashes and the technical challenges involved, that statement is almost certainly a bluff. It's a great example of someone using technical-sounding language to make an empty threat.
>As we covered, cryptographic hashes are completely unique to each image, and even the smallest change, like a single-pixel edit, creates a brand new hash. Someone would have to get the hash for every single AI image ever posted and then somehow get a new hash every time someone made the slightest change to their image, which is impossible.
>And if they're talking about a perceptual hash, the idea of creating a single filter to identify all AI-generated images is an incredibly difficult technical problem. It would require building a sophisticated machine learning model, which is something that even the companies that make these AI tools are still perfecting.
>So, for a random person to credibly make such a filter and then continuously update it to block new images and new editing techniques is not feasible. The threat to "force" people to "censor yourself and artifact every dogshit image" is completely hollow.
>That kind of talk is just noise. He sounds like an idiot. He's talking a big game without understanding the basic technology. Trying to create a universal AI image filter is like trying to catch every grain of sand with a single sieve—it's just not going to happen.