4 ms·
Considering how frequently my innocent images (petting my cat, taking a photo of a pepper growing in my garden, pointing to a word printed on a page in a book)
by err4nt 5y ago
Considering how frequently my innocent images (petting my cat, taking a photo of a pepper growing in my garden, pointing to a word printed on a page in a book) trip Discord's automatic-bad-image-blocker and are unable to be uploaded to Discord because they think they are bad, I'm a little worried that when they turn this technology on, instantly many 100% innocent photos people have taken will be quietly flagged and put people on lists without them ever knowing it, and as they go through life this non-thing that they were never charged with (to be able to be found innocent) will follow them and haunt them and impact other business relationships, etc. All because of pictures a computer thought was something else.
- nmeofthestate 5y agoSounds like you're describing a completely different system than the one being discussed. It compares to a corpus of known-bad images.
- cwkoss 5y agoThe contents of this corpus are unverifiable. They are claimed to be only known-bad images. However, the public doesn't (and shouldn't) have the ability to inspect these images to verify that non-CSAM politically threatening images are not included. (ex. pictures of politicians doing cocaine with adult prostitutes, or images of war crimes committed by US troops) I think that distinction is important. Because even if the NCMEC database is only child porn right now (which we cannot know), the public will also not know if the scope of the database later changes because now Apple will call the police on whoever possesses an image that matches this database. Adding this capability increases the incentive for misuse of the database.