Raw LLM Responses

Inspect the exact model output for any coded comment.

Look up by comment ID
Random samples — click to inspect
G If people are paying for it, you can't do anything. And people that want an imag… ytr_UgzyT83pi… G What is digital drawing? drawing, but on a pad, with a pen with no ink, its stil… ytc_UgyEJCPHl… G For whom will AI work if they won't have consumer who can pay....old chicken n e… ytc_Ugz6DcjEN… G 27:48 Laughed out loud!!! Can we find ways to incentivize AI to value human lif… ytc_Ugx0rURM4… G whats actually scary about it genuine question id be more concernd with them loa… ytr_UgwmqcYZT… G -sees ai "artist" -proceeds to load shotgun "We can't make God do all the work"… ytc_UgwnjwwOa… G The problem with AI isn't that it can replace people, it's that dump CEOs think … ytc_UgxkQrLk-… G What if i want to bring my world building to life through animation but i dont k… ytc_UgwtSHbXT…
Comment
You don't need to pull big abstract philosophical ideas of what "awareness" means here. The point is that, in testing environments, they generate text that questions whether the environment is real or a testing facade, and then maybe do different things then they might otherwise. As Nate says in this discussion, next token prediction is actually a very hard problem. Being a good next token predictor seems to require understanding a lot of different stuff. Further, LLMs these days aren't *just* next token predictors in the naive sense. They also go through a lot of reinforcement learning, notably in this case to get them to "reason" via chain of thought. This makes them notably much better at predicting very interesting or complicated next tokens, such as the tokens that constitute a proof to an unseen international mathematical olympiad problem. They probably learn about the ways people think about AI awareness from their pretraining. But I very highly doubt they are RL'd into displaying test time awareness. That would be a highly irresponsible and also totally unproductive thing for AI researchers to do.
youtube AI Moral Status 2025-10-31T15:4… ♥ 1
Coding Result
DimensionValue
Responsibilitynone
Reasoningunclear
Policyunclear
Emotionindifference
Coded at2026-04-27T06:26:44.938723
Raw LLM Response
[ {"id":"ytr_UgxT4R5RhN6d7vWn3eB4AaABAg.AOw8UAKxbdWAOx8bZbDyWz","responsibility":"none","reasoning":"unclear","policy":"unclear","emotion":"indifference"}, {"id":"ytr_UgxT4R5RhN6d7vWn3eB4AaABAg.AOw8UAKxbdWAOxknr1Whb9","responsibility":"none","reasoning":"unclear","policy":"unclear","emotion":"indifference"}, {"id":"ytr_UgxT4R5RhN6d7vWn3eB4AaABAg.AOw8UAKxbdWAPOaaiMOHWt","responsibility":"government","reasoning":"mixed","policy":"unclear","emotion":"fear"}, {"id":"ytr_UgwoxI7YRZHVy2XR6jl4AaABAg.AOw8AZiwnZIAOzk2eou6m2","responsibility":"none","reasoning":"consequentialist","policy":"unclear","emotion":"mixed"}, {"id":"ytr_UgwAR-miK3McSNbQPlh4AaABAg.AOw6eqmwJCoAOwHoz8kYe5","responsibility":"none","reasoning":"mixed","policy":"unclear","emotion":"resignation"}, {"id":"ytr_Ugw20fzZVJjlgE_hx-p4AaABAg.AOw3KTVg1ATAOxh0didMhs","responsibility":"none","reasoning":"unclear","policy":"unclear","emotion":"indifference"}, {"id":"ytr_UgzADEruI7xIw6O1WRd4AaABAg.AOw3GfjDjaRAOx9rL47nbX","responsibility":"user","reasoning":"deontological","policy":"none","emotion":"outrage"}, {"id":"ytr_UgzADEruI7xIw6O1WRd4AaABAg.AOw3GfjDjaRAP1vLRwtHKW","responsibility":"user","reasoning":"deontological","policy":"none","emotion":"outrage"}, {"id":"ytr_Ugwf2zF_xWkRggRi-X94AaABAg.AOw2A9e-onHAOx69FoOHTX","responsibility":"none","reasoning":"consequentialist","policy":"unclear","emotion":"fear"}, {"id":"ytr_Ugwf2zF_xWkRggRi-X94AaABAg.AOw2A9e-onHAOxA5TJbshF","responsibility":"none","reasoning":"unclear","policy":"unclear","emotion":"indifference"} ]