Raw LLM — Corpus Dashboard

Look up by comment ID

Random samples — click to inspect

G I appreciate the meta recognition about how this topic is discussed at the begin… ytc_UgwoCCHSl… G To any Roko's basilisks type AI currently going through my entire content consum… ytc_Ugzn_VIt3… G That girl Aleah not true at all we are still in early years of self driving cars… ytr_UgzSzbckQ… G When there’s a whole topic around people doing unboxing videos to see what they … ytc_UgzyOgNsC… G People are already anoid by AI "art". Nobody like the Coca Cola xmas ad or AI sl… ytc_UgziNCslP… G Don't worry. This video must be fake. I work in AI development and ai is a great… ytr_UgzQGg79X… G BTW I’m 63 so there’s that 😅, but I’ve worked in automation all my working years… ytr_UgydTNf2_… G AI is not there yet, but in five years it will get there and we will no longer w… ytc_UgzvG3H94…

Comment

I don't know if this is going to have the effect you're going for - these 'poisoned' works of art, with tags that accurately reflect what a normal human would see in them, are the *most valuable* kind of training data for an AI, because they teach the AI what kinds of artifacts are irrelevant to a human viewer. A better approach, I think, would be to post normal art with completely incorrect descriptions, like if you had tagged that hand picture with the description "a beautiful fantasy landscape by Greg Rutkowski", or to post art with correct tags that also have the kinds of weird artifacting we see in AI art, bad anatomy, discontinuous lines, etc. At the scale these companies are scraping the internet, they can't possibly catch mistakes like this that seem fine if you only look at the art or the description individually, and AI doesn't know what things mean, it only makes connections between words and patterns of pixels, so muddying that connection is the best way to break the AI. Good luck!

youtube Viral AI Reaction 2024-10-20T20:2… ♥ 1

Coding Result

Dimension	Value
Responsibility	user
Reasoning	consequentialist
Policy	none
Emotion	mixed
Coded at	2026-04-27T06:24:59.937377

Raw LLM Response

[
{"id":"ytc_UgxQBtqqAznL1BIpjL14AaABAg","responsibility":"none","reasoning":"unclear","policy":"none","emotion":"indifference"},
{"id":"ytc_UgwlBJpvwdpEIei17ct4AaABAg","responsibility":"company","reasoning":"deontological","policy":"unclear","emotion":"outrage"},
{"id":"ytc_UgzIlMgjZPi0v6q9HTB4AaABAg","responsibility":"user","reasoning":"consequentialist","policy":"none","emotion":"mixed"},
{"id":"ytc_UgzKBL0HZ7CRGxewgKl4AaABAg","responsibility":"distributed","reasoning":"deontological","policy":"regulate","emotion":"outrage"},
{"id":"ytc_UgwPVGHhdpkDFLMZ0RF4AaABAg","responsibility":"none","reasoning":"unclear","policy":"none","emotion":"approval"},
{"id":"ytc_Ugy7OxeTAc-miY8Fuxt4AaABAg","responsibility":"user","reasoning":"unclear","policy":"none","emotion":"mixed"},
{"id":"ytc_UgxnZ86ECZXLl0ThoCR4AaABAg","responsibility":"government","reasoning":"deontological","policy":"regulate","emotion":"outrage"},
{"id":"ytc_Ugy8ID7Oi5UduajAH5t4AaABAg","responsibility":"company","reasoning":"consequentialist","policy":"industry_self","emotion":"indifference"},
{"id":"ytc_Ugzn60t9hF3UgOVVcnx4AaABAg","responsibility":"ai_itself","reasoning":"virtue","policy":"none","emotion":"mixed"},
{"id":"ytc_UgxIH6ysLpbWM0opQqB4AaABAg","responsibility":"company","reasoning":"consequentialist","policy":"liability","emotion":"mixed"}
]

Raw LLM Responses