No! Thanks for providing a valid test 😂
Discussion
You’re welcome 🤗🙃🧡
I don't think there's anyone better to train a NSFW model on than nostr:npub12z8jsett3k6rv9fa2guau5p540qr2xuvjzkr8e432mglafjt99sqkw9zmx. She's the poster woman.
Also consistent with tagging her posts.
Unfortunately the problem is mostly false negatives, not positives :(.
I think I’ll probably just replace this model with something more robust/accurate.
Example: it classified this nostr:npub1qqqqqqyz0la2jjl752yv8h7wgs3v098mh9nztd4nr6gynaef6uqqt0n47m photo as NSFW, I guess because of the leg 😂

Bahaha 🤣🥲
Too sexy, apparently.
{
"data": {
"drawings": 0.07919999999999999,
"hentai": 0.0068,
"neutral": 10.9955,
"porn": 0.3348,
"sexy": 88.5837,
"is_nsfw": true
}
}
I understand.
Just out of curiosity:
Are they usually false negatives for my posts? Or are you speaking in general with AI bots, or other n s f w accounts ?
Generic other posts. I also have it automatically mark NSFW if the content is tagged (like you always do).
Have you considered applying a Graph Neutral Network to lend to this? Followers of Onyx would be more likely followers of other NSFW content and could be used as an indicator of the nature of the content.
Cool idea. Right now mostly focused on topic classification of individual notes but this type of thing could be useful for grouping users together by common interests.
I have more ideas than time unfortunately 😂
What is positive and what is negative?
I hate the idea that we have to cater to prudes. If there is a negative reaction does my visibility for other posts go down?
People can block naked people if they want to but I despise the concept that it needs to be hidden. The nudity is natural. How people portray it can be unnatural or illegal.
Go after them with the detectors, not us working legally from home.
It's like Jack Mallers catering to Visa Sesta Fosta.
This is just sentiment analysis from a language model.
None of this affects your posts, visibility, or anything else. It has nothing to do with clients or relays. I’m simply experimenting with topic classification of all nostr posts.
Thank you. 🙏 I am honoured you’d say such a thing. 💜🦄🥹☺️