Raw LLM — Corpus Dashboard

Look up by comment ID

Random samples — click to inspect

G I'll believe it when AI can collapse the wave function while observing the 2 sli… ytc_UgwLRlzVT… G People always say that someone using Ai will replace US not ai but it does'nt ma… ytr_UgzcDPeK6… G I can also confirm that chatGPT is 100% woke feminist. It told me that men gett… ytc_UgxSrWXNF… G How do you feel about the use of artificial intelligence in political ads? Do yo… ytc_Ugxn6S3Ys… G I hate to say it but the market has become more important to network, then build… rdc_n96apnx G All of these Ai guys and everyone else in the game always say a "Super Intellige… ytc_Ugybp0Ov0… G People are a huge liability, in professions like truck driving. Plus it's a bori… ytc_Ugx6zCc5L… G If all the jobs are replaced by AI? Who will pay for all the products AI make?..… ytc_Ugy8bDdDZ…

Comment

16:09 not exactly. That's also part of it, but the important point is that saying "I don't know" isn't a "good answer" that's useful to us as users. The LLM does not know if it knows things. That is the problem. So it has no way to say internally, ah I'm not confident here so I should be less overt in phrasing, or vice versa, ah there's tons of evidence here so I'll be very clear. It just roleplays as though it always has the answer, because a machine that has all the answers, IS WHAT WE'RE ASKING IT TO BE. If you try to train an LLM to be less certain, it's just going to go "uhh idk" for EVERY ANSWER, because that's the only answer that would be true. And early versions of GPT, particularly 3.5, basically did that. "As an AI language model..." Otherwise, it will be "uncertain" when the common presentation of an issue is one of uncertainty, or when the previous context inspires it to think that phrasing is likely. And it will do that even when the answer is super obvious and it definitely has that answer in the training set. It has no way to tell itself when to do it one way or the other. It only knows that we want it to sound like it's saying things that are true.

youtube AI Moral Status 2026-01-08T16:1… ♥ 1

Coding Result

Dimension	Value
Responsibility	developer
Reasoning	deontological
Policy	liability
Emotion	mixed
Coded at	2026-04-26T23:09:12.988011

Raw LLM Response

[
{"id":"ytc_UgxXvP06xB_rvHXU8nl4AaABAg","responsibility":"ai_itself","reasoning":"consequentialist","policy":"unclear","emotion":"fear"},
{"id":"ytc_UgxB2lUMC10V2WCKMdh4AaABAg","responsibility":"government","reasoning":"consequentialist","policy":"regulate","emotion":"fear"},
{"id":"ytc_UgzG6m5nNk-ZQp4yPdd4AaABAg","responsibility":"government","reasoning":"consequentialist","policy":"regulate","emotion":"fear"},
{"id":"ytc_UgzdLgUpm0zqRww_36x4AaABAg","responsibility":"none","reasoning":"consequentialist","policy":"none","emotion":"resignation"},
{"id":"ytc_UgxXKB0Q9EOyb0TYAQ54AaABAg","responsibility":"developer","reasoning":"deontological","policy":"liability","emotion":"outrage"},
{"id":"ytc_Ugz9jWegCqJ5MLH9GXF4AaABAg","responsibility":"unclear","reasoning":"unclear","policy":"unclear","emotion":"indifference"},
{"id":"ytc_UgxjHKweqa7s6ZC0JHB4AaABAg","responsibility":"ai_itself","reasoning":"consequentialist","policy":"ban","emotion":"fear"},
{"id":"ytc_Ugy-KZ4-7G2BKQOny894AaABAg","responsibility":"developer","reasoning":"deontological","policy":"liability","emotion":"mixed"},
{"id":"ytc_Ugy88yz9_C5B-z5vALJ4AaABAg","responsibility":"distributed","reasoning":"consequentialist","policy":"regulate","emotion":"fear"},
{"id":"ytc_Ugx-G5YAEcxVcUZLiZt4AaABAg","responsibility":"unclear","reasoning":"mixed","policy":"unclear","emotion":"mixed"}
]

Raw LLM Responses