Raw LLM Responses

Inspect the exact model output for any coded comment.

Look up by comment ID
Random samples — click to inspect
G I’m pretty anti-AI but I wasn’t feeling great and therapy was making things wors… ytc_UgzmsY8bx… G We need to start developing the equivalent of nuclear weapon but fires EMP. The … ytc_UgxsPwaRn… G Ai also likes to suggest suicide to its users so this really isn’t surprising. T… rdc_o7eymaq G The worst part is when you see some artwork that looks really nice at first glan… ytc_UgzLCU2RI… G Hope times wins so other companies can sue the shit out of these ai companies ti… ytc_UgxEziN-E… G I have tried quitting using AI on two occasions But it never lasted 😞 I just c… ytc_UgyK8r3wS… G Just a little more Alex, just a little more, I think maybe to confuse him with s… ytc_UgyfjKDZy… G She said NOOO, HELP ME!!! AAGGHHH UMMM, WTF!!! THIS IS BEYOND CREEPY????¡!!!!!!… ytc_UgyUJBCGJ…
Comment
They state that they gave the instruction "the code only" but GPT decided to add the markdown quotes and additional explanations on multiple occasions? This does imply that GPT is less compliant with requests and does not follow the users prompts as closely. I don't see how "their tests don't work" here lol. Yeah the code might be right, but if (like they say in the paper) you use chatGPT as a code generator in a pipeline, you now have to have an additional parser that checks for python markdown and other text and deletes it. Is this a minor change that does not change code quality? Probably yes. Does it decrease performance in a production environment? Absolutely. The test is valid. You just dismiss these changes because they do not seem important to you and your use of ChatGPT. edit: typo
reddit AI Harm Incident 1689776098.0 ♥ 12
Coding Result
DimensionValue
Responsibilityai_itself
Reasoningconsequentialist
Policynone
Emotionindifference
Coded at2026-04-25T08:33:43.502452
Raw LLM Response
[{"id":"rdc_jslc385","responsibility":"ai_itself","reasoning":"consequentialist","policy":"none","emotion":"indifference"},{"id":"rdc_jslo95t","responsibility":"none","reasoning":"deontological","policy":"none","emotion":"outrage"},{"id":"rdc_jsm370w","responsibility":"company","reasoning":"consequentialist","policy":"none","emotion":"frustration"},{"id":"rdc_jsmea4a","responsibility":"company","reasoning":"virtue","policy":"none","emotion":"outrage"},{"id":"rdc_jsk7i6o","responsibility":"none","reasoning":"consequentialist","policy":"none","emotion":"indifference"}]