Raw LLM Responses
Inspect the exact model output for any coded comment.
Look up by comment ID
Random samples — click to inspect
G
when AI became available, people was thinking about recording meetings and let A…
ytc_Ugx4Twukd…
G
I find it very funny when tech bros fantasize about ai becoming a evil super int…
ytc_UgzldzcYL…
G
At least not knowingly.... it's very important that was in there because they k…
ytc_UgyejSht2…
G
Hey Hank. Yudkowsky and "AI in Context" come from a long line of Bay Area sci-fi…
ytc_Ugyx5RFwQ…
G
9:40 🚨🚦📣😳😬 wow AI machines now debating about human nature?. AI machines even te…
ytc_UgygQL0JY…
G
Well, the buggest reason why I could see it be a bad thing is if humans rely too…
ytc_Ugwv0lIge…
G
It's obvious that when AI has the "power" and knows the "nature of human beings,…
ytc_UgwjOI-o5…
G
We are human when we see ourselves in risk we destroy it. do not worry nothing c…
ytc_Ugxn8KXxj…
Comment
One idea to keep in mind is that you can use a cheap AI model to augment GPT-4/5 or even human output.
A joke example is replacing the word "wand" with "wang" in the Harry Potter stories. Taping knives to roombas. Or consider how not every employee was aware they were working on the atom bomb (or are working at scam organizations today). Basically, advanced jailbreaking, as opposed to those jailbreaks that should be obvious to fix.
I don't know if such a technique would actually scale for truly dangerous scenarios, but I believe it'd definitely scale for hate speech and erotica, and I've already found some success with this technique with barely any postprocessing at all. OpenAI would also probably not really care about this kind of misuse, so long as they weren't directly responsible.
Terrorist level misuse is a different story, and I'm not sure how you could avoid the possibility without severely handicapping your product. Considering helpful business emails and manipulative phishing scams are basically identical, as one example...
reddit
AI Responsibility
1682548472.0
♥ 2
Coding Result
| Dimension | Value |
|---|---|
| Responsibility | none |
| Reasoning | unclear |
| Policy | none |
| Emotion | mixed |
| Coded at | 2026-04-25T08:33:43.502452 |
Raw LLM Response
[{"id":"rdc_jhspuqw","responsibility":"none","reasoning":"unclear","policy":"none","emotion":"indifference"},{"id":"rdc_jht26c9","responsibility":"none","reasoning":"unclear","policy":"none","emotion":"indifference"},{"id":"rdc_jhsqwc5","responsibility":"none","reasoning":"unclear","policy":"none","emotion":"indifference"},{"id":"rdc_jhsre0c","responsibility":"none","reasoning":"unclear","policy":"none","emotion":"approval"},{"id":"rdc_jhuh106","responsibility":"none","reasoning":"unclear","policy":"none","emotion":"mixed"}]