Raw LLM Responses
Inspect the exact model output for any coded comment.
Look up by comment ID
Random samples — click to inspect
G
Listening to this is the car when I went thru a drive thru and AI took my order …
ytc_UgyHhUJP3…
G
Um Gemini has everything and ChatGPT is missing most things. This is the day as …
ytc_UgwJZZsGn…
G
impossible dans le BTP, par contre l'IA pourrait être une aide surtout au niveau…
ytc_UgxCt98Cx…
G
"hello? I need to report an ai"
"Sir this is the fire fighter."
"Use the water h…
ytc_Ugz-eypdk…
G
You are explaining mostly supervised learning which is a basic form of learning …
ytc_UgwNlWSHY…
G
I work in development, and AI is quite useful.... it probably saves me about 30 …
ytc_UgxnuDWs3…
G
That fear is real and not irrational. But there's a difference between "eventual…
ytr_UgyKMJzAT…
G
A.I. Is the artist, Prompters are commissioners.
Credit of the art should go to …
ytc_UgyGZbC5k…
Comment
Ok, I went back and read the Nightshade paper (Shan et al), and the authors know what they're doing, it's actually pretty cool - basically, modern AIs work by converting collections of pixels into a simplified 'latent space' where the concepts behind the art are grouped together as detected by a different image-to-text AI model, and Nightshade works by finding a way to trick that model into putting the image into completely the wrong place in the latent space, so conceptually it's doing something similar to what I suggested mislabeling the art but much more efficiently and effectively - against this particular AI. In the long term, though, I still think that if you trained a new image-to-text model from scratch using this data, which to be fair none of these companies are doing because it'd be way too expensive, then these poisoned images would be valuable in the way I assumed they would be before.
youtube
Viral AI Reaction
2024-10-20T21:2…
Coding Result
| Dimension | Value |
|---|---|
| Responsibility | none |
| Reasoning | unclear |
| Policy | none |
| Emotion | approval |
| Coded at | 2026-04-27T06:24:53.388235 |
Raw LLM Response
[
{"id":"ytr_UgwjgT2xgHcxGJIkDRV4AaABAg.A9pVUyKSjzbAA4P1oZS8cl","responsibility":"none","reasoning":"unclear","policy":"none","emotion":"indifference"},
{"id":"ytr_UgwxH1gLqg1lpbIx9yJ4AaABAg.A9pVHjb2kU9A9q-FeUv4ds","responsibility":"ai_itself","reasoning":"virtue","policy":"none","emotion":"outrage"},
{"id":"ytr_UgwxH1gLqg1lpbIx9yJ4AaABAg.A9pVHjb2kU9A9qfSoLdpC_","responsibility":"developer","reasoning":"virtue","policy":"none","emotion":"indifference"},
{"id":"ytr_UgwxH1gLqg1lpbIx9yJ4AaABAg.A9pVHjb2kU9A9rQsNgzoBq","responsibility":"none","reasoning":"unclear","policy":"none","emotion":"indifference"},
{"id":"ytr_UgzIlMgjZPi0v6q9HTB4AaABAg.A9pV5dxGYTMA9pbEU1SPxq","responsibility":"none","reasoning":"unclear","policy":"none","emotion":"approval"},
{"id":"ytr_Ugx3xiS8IcOSsNDplwl4AaABAg.A9pUyNe5IV5A9p_KVWfh1x","responsibility":"none","reasoning":"unclear","policy":"none","emotion":"indifference"},
{"id":"ytr_Ugx3xiS8IcOSsNDplwl4AaABAg.A9pUyNe5IV5A9pbLeixia2","responsibility":"company","reasoning":"consequentialist","policy":"regulate","emotion":"outrage"},
{"id":"ytr_Ugx3xiS8IcOSsNDplwl4AaABAg.A9pUyNe5IV5A9pdhC79g-x","responsibility":"developer","reasoning":"deontological","policy":"none","emotion":"outrage"},
{"id":"ytr_Ugz6QScWr6E3XxX7F5J4AaABAg.A9pUwgo0NfFAG_u0vQIG9r","responsibility":"none","reasoning":"unclear","policy":"none","emotion":"outrage"},
{"id":"ytr_Ugw1VVFhwGjEeEskUGV4AaABAg.A9pUc8wywZCA9pjcgmb8PG","responsibility":"government","reasoning":"consequentialist","policy":"ban","emotion":"outrage"}
]