←back to thread

534 points BlueFalconHD | 3 comments | | HN request time: 0s | source

I managed to reverse engineer the encryption (refered to as “Obfuscation” in the framework) responsible for managing the safety filters of Apple Intelligence models. I have extracted them into a repository. I encourage you to take a look around.
1. michaelt ◴[] No.44484528[source]
I assume all the corporate GenAI models have blocks for "photorealistic image of <politician name> being arrested", "<politician name> waving ISIS flag", "<politician name> punching baby" and suchlike.
replies(2): >>44484622 #>>44484876 #
2. lupire ◴[] No.44484622[source]
Maybe so, but think about how such a thing would be technically implemented, and how it would lead to false positives and false negatives, and what the consequences would be.
3. bigyabai ◴[] No.44484876[source]
Particularly the models owned by CEOs who suck-up to authoritarianism, one could imagine.