Urusan on Nostr: AI augmented moderation would actually be pretty easy to put together, though you ...
AI augmented moderation would actually be pretty easy to put together, though you probably shouldn't give it the banhammer directly and just have it report to you.
Set up a LLM, Llama 3 8B based models should perform well, then craft a standard prompt explaining what you are looking for in detail. It can be quite long to nail down all the details.
In your prompt, include a specification for a machine readable classification string that can be extracted via script. Then run it on posts
Published at
2024-05-10 06:52:45Event JSON
{
"id": "b7d3363a0710c419fdb4c60c2c48a64e6cfb1582105ed961ec9fb280ea81c43c",
"pubkey": "f3f28a08e7c4ffccb47484ae2d4b6f567ec5621de6eba00aa82d627891cdaf8e",
"created_at": 1715323965,
"kind": 1,
"tags": [
[
"p",
"8240035a9f30abf3fd91387f32e922b6a1240df57fb21d9b267c1c87c65434da"
],
[
"e",
"4b458d51e6e3dea69c86095adc53aa20b01a3c6524341993aeebcafc634972e3",
"",
"root"
],
[
"proxy",
"https://aipub.social/@urusan/112415471417410936",
"web"
],
[
"proxy",
"https://aipub.social/users/urusan/statuses/112415471417410936",
"activitypub"
],
[
"L",
"pink.momostr"
],
[
"l",
"pink.momostr.activitypub:https://aipub.social/users/urusan/statuses/112415471417410936",
"pink.momostr"
]
],
"content": "AI augmented moderation would actually be pretty easy to put together, though you probably shouldn't give it the banhammer directly and just have it report to you.\n\nSet up a LLM, Llama 3 8B based models should perform well, then craft a standard prompt explaining what you are looking for in detail. It can be quite long to nail down all the details.\n\nIn your prompt, include a specification for a machine readable classification string that can be extracted via script. Then run it on posts",
"sig": "614024c272e69d0ca5339a84ec87b6c91d2a87b5e2350c2e5184519770457c569fc7bbf61e8e609f1da17b98c88ce6b5445c8e33614ea2a6783352a093e42556"
}
