One advertisement featured a pornographic video with a deepfake closely resembling a prominent US politician. Apple removed the app from the App Store after an inquiry from WIRED.
Nudes classification predates LLMs by decades. Its accurate, and dirt cheap, because it had to be to detect CP.
You can bet Lemmy.world uses it. Plenty of smaller companies do it. Pick any NSFW picture on CivitAI, and you can see how insanely accurate autotagging has gotten.
So, no… this was not an accident on Facebook’s part.
They’re pretending that they’re still in 2008 where a basic word or letter swap in the copy wouldn’t trigger the blocklist.
Facebook 100% knows about those ads and is willing to be complicit in the scams for pure profit.
With the massive datacenters running their useless LLMs there’s no way that they couldn’t detect those ads.
Nudes classification predates LLMs by decades. Its accurate, and dirt cheap, because it had to be to detect CP.
You can bet Lemmy.world uses it. Plenty of smaller companies do it. Pick any NSFW picture on CivitAI, and you can see how insanely accurate autotagging has gotten.
So, no… this was not an accident on Facebook’s part.