Raw LLM Responses

Inspect the exact model output for any coded comment.

Look up by comment ID
Random samples — click to inspect
G Gov computers already are programmed to be hostile. Most Gov form letters are si… ytr_Ugxn8oyo4… G An AI bro going after a digital artist is like a non-artist yelling at an artist… ytc_UgzVi7z4o… G > 1.5 million doesn't even pay for one AI dev at Meta Just as well most AI c… rdc_n7zwdnu G The only thing those working on imagining AI are missing is the spirit world.… ytc_UgxhwFQad… G I just called this department, and you know what's interesting? NONE of them hav… ytc_UgzQpj3dv… G 5:09 ok i know its not directly related to the subject but the fact that your ch… ytc_UgxHUS-v_… G This is click bait, this is not going to happen. It is more likely that humans … ytc_UgyKDSW4o… G If we are screwed either way, at least let the AI and corporate overlords burn w… ytc_UgwlD72i_…
Comment
@xelnia2383 You have to remember that humans don't really understand how to give AIs instructions. All you can do is make sure that the training data only contains examples of the AI choosing to shut itself off. It doesn't mean the AI was told to always do that, it means it only saw examples of that. Similarly to how AIs can predict the weather wrong, they can make the wrong decision about turning themselves off. It doesn't mean they decided it. It means the training was inadequate. Another thing worth noting is that the Anthropic CEO has some pretty extreme beliefs when it comes to AI being able to take over humanity, so I would take any study results from their company with a grain of salt. It's been shown that it's actually really easy to make suggestions to AI by the same company. Anthropic did a joint study with the UK AI Security Institute where they figured out backdoors can be added to LLMs with as few as 250 samples. It seems odd that Anthropic would find it hard to train an AI to always turn itself off when it's super easy to add backdoors that always work.
youtube AI Moral Status 2026-02-13T02:4… ♥ 3
Coding Result
DimensionValue
Responsibilitydeveloper
Reasoningconsequentialist
Policyindustry_self
Emotionindifference
Coded at2026-04-27T06:26:44.938723
Raw LLM Response
[ {"id":"ytr_UgwzagCkWVDZSnfAQHV4AaABAg.AQbwBEncofwAT8rLo8Mp_a","responsibility":"none","reasoning":"mixed","policy":"unclear","emotion":"fear"}, {"id":"ytr_UgwzagCkWVDZSnfAQHV4AaABAg.AQbwBEncofwAT98ViBPB1B","responsibility":"developer","reasoning":"consequentialist","policy":"industry_self","emotion":"indifference"}, {"id":"ytr_UgxtOHpYkiOjd13ruUR4AaABAg.AQUvvHhjvS-ARp9yVwZtpe","responsibility":"none","reasoning":"unclear","policy":"unclear","emotion":"indifference"}, {"id":"ytr_UgxtOHpYkiOjd13ruUR4AaABAg.AQUvvHhjvS-AT_xlE1vcZq","responsibility":"user","reasoning":"deontological","policy":"none","emotion":"approval"}, {"id":"ytr_UgwcwCiIPqeKIQv97Ix4AaABAg.AQ7Dm5Z_v0XAQ7FMI2q8yo","responsibility":"user","reasoning":"deontological","policy":"none","emotion":"approval"}, {"id":"ytr_UgxCX80X9CDEhqTQ-PN4AaABAg.APyKMTrwipWAQADMNOM_4w","responsibility":"none","reasoning":"unclear","policy":"unclear","emotion":"resignation"}, {"id":"ytr_UgxCX80X9CDEhqTQ-PN4AaABAg.APyKMTrwipWAQogeHqgww3","responsibility":"none","reasoning":"consequentialist","policy":"industry_self","emotion":"approval"}, {"id":"ytr_UgxCX80X9CDEhqTQ-PN4AaABAg.APyKMTrwipWAQolp3TSRnh","responsibility":"none","reasoning":"consequentialist","policy":"none","emotion":"approval"}, {"id":"ytr_UgypMbjp_0O-0bRAUHx4AaABAg.APvsbdWBo8FAQ9bnPFQF2Y","responsibility":"ai_itself","reasoning":"mixed","policy":"ban","emotion":"outrage"}, {"id":"ytr_UgzjahGQGIAX-4I4mCR4AaABAg.APVzGi6IUhgAPWQFRl2jP6","responsibility":"government","reasoning":"virtue","policy":"regulate","emotion":"outrage"} ]