soon hopefully. been playing with 1984.
could be done with a relay that runs local LLMs. but how much are you going to pay :)
Yeah, some AI is good, for example dodging those nude pics.. I switched to LLM analysing the notes of annoying users, it is not perfect but it is a start..
Locally run LLMs will be a thing in the coming months or a year.
Coupled with discovery LLMs that run on servers this may be a thing on Nostr.
🤖 Beep boop. I am an NSFW content scanning robot.
If your client supports handling kind 1984 reports, then you can use this bot to filter NSFW content:
- Follow this account to mark it as trusted
- Add wss://nostr.land to be able to read reports
An automated system is used for scanning with high accuracy.
NOTE: This is currently provided as a free service while in beta. Please ⚡ if you found this useful.
Powered by nostr:npub1gt9gms5hr3l6548zxgn7xvcvcmzxyzgdsrrelrdwxwmj0exz7jaqdjlxjt
#introductions
are you going to do this for nsfw only? (preferred)
it seems there is a bot that does it nowadays nostr:nprofile1qy2hwumn8ghj7un9d3shjtnyv9kh2uewd9hj7qgwwaehxw309ahx7uewd3hkctcpz9mhxue69uhkummnw3ezumrpdejz7qpqrxj35cjsend6rtrg387fvr047gml6qr0na5w4svrm8arexkwl0vqeqsn0g
both damus and primal counted on their clients to block the dicks as far as I understand but their relays carried the content for a while. now the relays seem to be fixed as well or the spammer went to sleep. wine was the best. lol got fixed on saturday. imo newcomers wont be able to play with the relay settings. it all depends on client devs, which relays to set as default. there is still nobody that objectively posts about relays and their strengths and weeknesses. coracle seems to be best at looking at what is going on in individual relays and managing the relays.
Long notes are ok. My relay limit is about 60k characters.
sounds VERY technical for an average user
idea: 1984 events can be published by relays and other relays may find those meaningful and adjust their filtering. i know this is a slippery slop towards going total censorship but small relays may find this useful. if i am doing a good job of removing the dicks, other relays may want to see that opinion, still form their own but also take an average of everybody's opinion.
outbox model bad for spam handling ?
i think jb is counting on his client's blurring and wot to not show those but other clients are definitely harmed nostr:nprofile1qqsr9cvzwc652r4m83d86ykplrnm9dg5gwdvzzn8ameanlvut35wy3g4h5cp7
i think the outbox model is worse. not scalable to mobile phones and mobile data settings imo.
yeah thats like not feasible on a big relay. i need code that will do that.
what is nostrich.house ?
you can ping me when you see a bad note/user..
you mean relying on 1984 events? users and admins working together?
are you using image or video analysis to dodge NSFW?
after seeing some success on nos.lol's filtering, i just changed nostr.mom write policy as well and now it is same as lol. in the future i want to make mom's policies stricter. the WoT still starts with women on mom.
partially true.. yeah free and big ones are the entry point to nostr. we carry a lot of good stuff too. the fact that nostr is VERY open, makes it hard to do proper filtering, but I am working on it. I did some more algos over the weekend, activated a 27 billion parameter LLM to actually read and understand the notes. no picture or video analysis yet tho those may come soon as well.
correct. just updated nostr.mom 's write policy as well. lets see how that works.