The f***ed up part is it actually f***ing works!
This is so cool, like, the Trump letter, and various things, they match up, and then, they end up with context.
Which is why I get pissed off, when some users, can evade detection, and put weird stuff in, when AI doesn't detect or catch it
I dunno it slips past
We're all frightened as platform operators because of 4chan CSAM stuff
4chan or actually Tumblr!
Tumblr*
They f***ed up
We're a publisher, so we are not as badly exposed, but we do work hard to detect weird stuff on our images and text