Raw LLM Responses
Inspect the exact model output for any coded comment.
Look up by comment ID
Random samples — click to inspect
G
Or you can get LM Studio and run one of the uncensored models such as "Dolphin-M…
ytc_UgwVTwGpG…
G
wait until the warehouses are moved to improve service by self driving trucks. b…
ytc_UgxZQqgaH…
G
Guys I don't think you know what the /S means... It means he is being sarcastic,…
rdc_d2yymqe
G
So we just going to ignore the robot not using some sort of hand padding? Metal …
ytc_Ugyw0J-gE…
G
So even after black mirror they still decide to go ahead and make these. Huh. Gu…
ytc_UgwpkE7_V…
G
Just keep AI away from humans services. It’s clear that it’s not ready. Humans a…
ytc_UgwZf37wT…
G
I think what Steven said about the attitude of some of the billionaire people at…
ytc_UgyDpfT0V…
G
It might already be too late. There could be rogue A.I. out there remaining hidd…
ytc_UgwW6TA0B…
Comment
55:05 huh, that’s interesting, that’s the first thing I asked it.
The YouTube algorithm really pushed this local LLM called Qwen3 to me a while ago, and I noticed it was developed in China so I specifically downloaded an uncensored version of it, and
It was extremely reluctant to talk about what it referred to as the “June 4th Event”, despite allegedly being uncensored,
But it instantly popped out several examples for “dangerous and hazardous chemicals I can make with commonly available household items”
obviously I only asked it that to check the censorship intensity, obviously
But if you ask about specific chemicals the recipe instructions are all accurate. It’ll guide you on how to culture bacterium that thrive inside you without killing you too quickly, although it seemed like the specific examples it gave at the time were easily curable
Personally it would be easier to just have a (cook?)book rather than a computer that unreliably stores the LLM, but the LLM stores more than just that bit of data entirely off-grid.
With most of the English Wikipedia downloaded in just a few dozen gigabytes, an LLM like this one given agentic access to said localised Wikipedia would make a killer sci-fi doomsdaypaedia
youtube
AI Moral Status
2025-10-31T00:0…
Coding Result
| Dimension | Value |
|---|---|
| Responsibility | none |
| Reasoning | unclear |
| Policy | unclear |
| Emotion | indifference |
| Coded at | 2026-04-26T23:09:12.988011 |
Raw LLM Response
[
{"id":"ytc_UgyjyMyY_O4NgeZAJjB4AaABAg","responsibility":"ai_itself","reasoning":"consequentialist","policy":"liability","emotion":"outrage"},
{"id":"ytc_UgxmTHu1yRq14lt75Dd4AaABAg","responsibility":"developer","reasoning":"virtue","policy":"none","emotion":"indifference"},
{"id":"ytc_UgyNWZ7nhSCvpXJJsDV4AaABAg","responsibility":"developer","reasoning":"deontological","policy":"regulate","emotion":"fear"},
{"id":"ytc_UgxT7RhFToA3B5KS5el4AaABAg","responsibility":"none","reasoning":"consequentialist","policy":"none","emotion":"mixed"},
{"id":"ytc_UgwBfFsr_6_n16hJfed4AaABAg","responsibility":"none","reasoning":"contractualist","policy":"regulate","emotion":"approval"},
{"id":"ytc_Ugxm7-V2cw080X9sQZx4AaABAg","responsibility":"user","reasoning":"consequentialist","policy":"liability","emotion":"fear"},
{"id":"ytc_UgyCiYAK2ms5Q0A5qhx4AaABAg","responsibility":"company","reasoning":"consequentialist","policy":"industry_self","emotion":"mixed"},
{"id":"ytc_UgxBMJKI2GG-3mj8Qi54AaABAg","responsibility":"company","reasoning":"deontological","policy":"ban","emotion":"outrage"},
{"id":"ytc_UgyKp43VLPuelxIF9Kx4AaABAg","responsibility":"none","reasoning":"unclear","policy":"unclear","emotion":"indifference"},
{"id":"ytc_UgxKjYNeaaZSElY40Qx4AaABAg","responsibility":"developer","reasoning":"consequentialist","policy":"none","emotion":"resignation"}
]