Raw LLM — Corpus Dashboard

Look up by comment ID

Random samples — click to inspect

G I hope that the deep fakes result in YouTube & Facebook start getting sued for p… ytc_Ugz3CyS02… G it can't, I'm using stable diffusion right now and let me tell u it doesn't gene… ytc_UgyKvPS8T… G السلام عليكم ورحمة الله وبركاته 🌸 I hope you ate doing well brother inshallah 😊… ytr_UgyJtm0iP… G I don't remember legacy cruise control getting this much scrutiny, despite it ha… ytc_Ugy16S_ez… G Its almost like the robot that cant draw a hand with 2.3 trillion dollars while … ytc_UgyrSGpoF… G Hey, I can wipe out the middle class. Well, Trump can help it along or vice vers… ytc_UgxkozUGE… G If 99% of jobs are gone who’s buying anything. At that point we don’t need AI or… ytc_Ugw9-hjaP… G Err... I'm a CG supervisor and can objectively say the 3d ice cream is way riche… ytc_UgwR2lzqp…

Comment

LLM and RAG systems hallucinate really badly from scientific sources. Somehow it keeps getting worse rather than better each time I run assessments. I hate it because it gets mixed in automatically when I am researching and it keeps f'ing me up. Do not use the general purpose tools for anything medical, and be real careful about the scientific. It makes sh*t up and then cites sources that do not contain what it makes up. It helps if you turn off access to the internet for internal RAG systems, but it still fs up if there isn't enough repeated information written in different ways. Information must be one topic per data source. No compare and contrast, no metaphors. Thankfully the systems built specifically for doctors work a lot better I'm told. I've seen enough B's in llms that it could be a human problem or it could be the system. Most people don't obsessively fact check multiple times for every single point. Llms context confuse frequently so taking info from chemistry and presenting as nutritional absolutely does happen.

youtube AI Harm Incident 2025-11-25T16:4…

Coding Result

Dimension	Value
Responsibility	company
Reasoning	consequentialist
Policy	liability
Emotion	outrage
Coded at	2026-04-27T06:24:53.388235

Raw LLM Response

[
{"id":"ytc_UgwiBUF0TkF7ynX_3bR4AaABAg","responsibility":"user","reasoning":"consequentialist","policy":"none","emotion":"indifference"},
{"id":"ytc_UgyGXcH9mby8-4hYqwl4AaABAg","responsibility":"distributed","reasoning":"consequentialist","policy":"unclear","emotion":"fear"},
{"id":"ytc_UgyU-RSLLQpl-nEJiAp4AaABAg","responsibility":"company","reasoning":"deontological","policy":"regulate","emotion":"outrage"},
{"id":"ytc_UgzUzg1e1D9UDCmiE9B4AaABAg","responsibility":"company","reasoning":"consequentialist","policy":"liability","emotion":"outrage"},
{"id":"ytc_UgykUh1RLKYbRB0lmw54AaABAg","responsibility":"none","reasoning":"unclear","policy":"none","emotion":"mixed"},
{"id":"ytc_UgwgWM_M2XaTwdzgb1d4AaABAg","responsibility":"distributed","reasoning":"mixed","policy":"unclear","emotion":"fear"},
{"id":"ytc_Ugxjx6V7LSQZJzWnwU14AaABAg","responsibility":"ai_itself","reasoning":"unclear","policy":"none","emotion":"indifference"},
{"id":"ytc_UgwrYHOCjSObdfqFDvl4AaABAg","responsibility":"company","reasoning":"deontological","policy":"ban","emotion":"approval"},
{"id":"ytc_UgzJguInZMTpcqbcj7N4AaABAg","responsibility":"ai_itself","reasoning":"unclear","policy":"none","emotion":"indifference"},
{"id":"ytc_UgzqCf9Pz6vptw4ugTN4AaABAg","responsibility":"user","reasoning":"consequentialist","policy":"industry_self","emotion":"resignation"}
]

Raw LLM Responses