#Primal you need to step up your game. Warning, extremely disturbing!
Discussion
nostr:npub1m2u4v2jtz4np0y3lk2anr4vm4vr4l7jlr7y94qglhfrgewsczchqg6y2gl this is actually pretty serious if you follow through on the linked content.
are their any good OSS filters out there using LLMs to prevent at relay? I image that would burn a lot of GPU/ASIC compute though
I don't know. I don't use Primal and the relays I'm using do a good job at filtering these kinds of things out. However, that post has been reportedly flagged by a bunch of people yet the content is still on their media server.
I guess that flagging and filtering out images is very expensive, because you need to use classic or ML based computer vision models. I cannot think to a cheaper way to do it. Plus, training a CNN to filter CSAM material means collecting such a dataset and training it. Training it means putting the images on a training environment...probably AWS or others, which would put you in a dangerous position...
Something may already exist out there, but certainly that is more complex than just filtering text.
LLMs are not the best tool for that probably
thanks. perhaps a simple blacklist service NIP could help. one that just stores the media and note references as text
That is a start, yes. Somewhere someone still have to recognize the malevolent content btw. Plus, if I set up script that loads into a relay thousands of copies of the same pic, every pic will have a different note reference, thus filters would have to store a large amount of note references for one pic.
Probably flagging accounts is way easier
I'd ideally like 2 settings to: 1)filter notes and direct media refs, and 2) also mute accounts as 2nd option. #dreams