Raw LLM Responses

Inspect the exact model output for any coded comment.

Look up by comment ID
Random samples — click to inspect
G Replace CEOs with a simple AI programmed for basic common sense and decency. We … ytc_Ugw8EhTh4… G The issue is that once we opened the AI box, there was no way to put it back in … ytc_UgyRQ_p8n… G We want ai to do the basics so we can do what we love not the other way round… ytc_UgwePf1nF… G Nothing wrong if AI art is only fed with their user's files and art, if an idiot… ytc_UgxBainds… G I'm with you on the topic of that AI art can be easy and simple to make and just… ytc_UgyawTqos… G Please spare us the automaton who parrot's what you place in it's programing to … ytc_Ugwh0hcUM… G So, what will happen? Will the corporations tell AI to "exterminate" the unempl… ytc_UgwcxuOcq… G Perhaps, would someone explain to me how vandalising Waymo vehicles contribute t… ytc_Ugw6L49An…
Comment
@mofumofutenngoku You are right that there was definitely alignment done, and can be understood by what topics it chooses not to speak about which we can tell by the fact that it was speaking about lot of those topics earlier, all that is alignment BUT the way it makes stories about whites and blacks is what you typically see on internet and I would argue you don't actually need to align chatgpt against whites to actually get it to make content like that. Remember that the prompts made in the test were neutral in nature so that brings out inner biases of the model which could both be added manually or could be present already in the data used. If it were added manually though it wouldn't give responses for questions which involve a bias from the user side which it does, for example if you asked about a black man joining a "friendly" white community then it will indeed give a proper response there and most likely a similar story to the white guy's. There has been a lot of jail breaking done and even the things that are added as safeguards to avoid sensitive topics have been circumvented and I've seen responses with all different kinds of biases which wouldn't be possible if the model itself was aligned internally. I might be wrong but afaik all the alignment I've seen from OpenAI side is limited to straight up disabling chatgpt from talking about certain topics in a negative way rather than actually coding in a bias with an agenda. If there is evidence for otherwise I'll definitely want a source.
youtube AI Bias 2023-10-13T02:5… ♥ 2
Coding Result
DimensionValue
Responsibilitydeveloper
Reasoningconsequentialist
Policynone
Emotionmixed
Coded at2026-04-27T06:26:44.938723
Raw LLM Response
[ {"id":"ytr_UgwmbuTwTbzhTlusp6x4AaABAg.9w850i5PEK39wPUZGCNRcU","responsibility":"none","reasoning":"unclear","policy":"none","emotion":"approval"}, {"id":"ytr_UgwB42trxGOjIqTkCe94AaABAg.9vv_mF_F45C9wFXqNb_rHE","responsibility":"developer","reasoning":"deontological","policy":"liability","emotion":"outrage"}, {"id":"ytr_UgwvU4-OAzgeHR5xvNx4AaABAg.9vqojy7qeQ5AIcwFDqEwtE","responsibility":"none","reasoning":"consequentialist","policy":"none","emotion":"indifference"}, {"id":"ytr_UgwcMf_YJmxEk9bf9EN4AaABAg.9vjiuk1Bdzw9vmqch_M21j","responsibility":"developer","reasoning":"consequentialist","policy":"none","emotion":"mixed"}, {"id":"ytr_UgwcMf_YJmxEk9bf9EN4AaABAg.9vjiuk1Bdzw9vnAlxLqOrs","responsibility":"developer","reasoning":"consequentialist","policy":"none","emotion":"mixed"}, {"id":"ytr_UgwcMf_YJmxEk9bf9EN4AaABAg.9vjiuk1Bdzw9wI2wGSCNDo","responsibility":"developer","reasoning":"deontological","policy":"liability","emotion":"outrage"}, {"id":"ytr_UgypqsxqTNjLQAo-bq54AaABAg.9DOaPHY-6BH9DOdH455zOt","responsibility":"none","reasoning":"unclear","policy":"none","emotion":"approval"}, {"id":"ytr_Ugycd9SlpC2nIiSmjx54AaABAg.9CqoKcL1i6o9E39125XaXj","responsibility":"none","reasoning":"unclear","policy":"none","emotion":"indifference"}, {"id":"ytr_UgyAPhayM4MvdkNVKy94AaABAg.9Cm75hQyt4_9D9bG6kARmL","responsibility":"none","reasoning":"unclear","policy":"none","emotion":"mixed"}, {"id":"ytr_UgwlH1oOqVf-ME5hRHd4AaABAg.9CiDdaSvRPs9ClBcbLGNja","responsibility":"user","reasoning":"consequentialist","policy":"none","emotion":"indifference"} ]