Raw LLM — Corpus Dashboard

Look up by comment ID

Random samples — click to inspect

G My problem with it is that people are taking artists work and pumping it into th… ytc_UgwmCQopS… G >Lee Se-dol strikes back to beat Google's DeepMind AI for first time **... a… rdc_d0z52tp G I see their robotics division has moved from the silicon valley to the uncanny v… ytc_UghOeUpr1… G Actually this robot can swim too. It was designed to be able to move on the grou… ytc_Ugxds5EG7… G So what is the end result? Could implants in human brains be the gateway of the … ytc_Ugw87EK6Y… G A hot dog is a cooked sausage so the argument is moot. Unless the hotdog is plac… ytc_UgypmVDoX… G It’s okay the ETs are coming out and AI will be least of your issues… ytc_UgwSEUtdA… G Senator Bernie Sanders, hello, we agreed on what you're telling us, but most of … ytc_UgypTS4sv…

Comment

"ust based on opinion, fails to construct his opinions ground up from first principles" Funny that, because all of the AI-safety results and discussion are the fruit of first-principles thinking he did in the early 2000s. Are you familiar with the inner/outer alignment? Mesaoptimizers? Instrumental convergence? At what level will the argument make sense to you and not sound like opinion? I'm paraphrasing some of the safety arguments: - we have ample evidence for the outcomes of adversarial agents where there is a dramatic asymmetry in their power - we have research to show that instrumental goals will develop, which will be aligned with the outer goals in the training regime but, due to distribution shift, will be non-aligned in the production regime - we have barely any mechanistic interpretability to make those inner goals obvious to us - in agentic systems, inner goals will be taken to extremes because they are intrinsically "desired" (were correlated with reward in training) - we expect instrumental convergent goals like self-defense, resource accumulation, and deception - we have near certainty that AI progress will continue at pace - we have near certainty that our safety mechanisms are not sufficiently in place to bear the live stress test of a superintelligence

youtube AI Governance 2024-12-27T03:1… ♥ 4

Coding Result

Dimension	Value
Responsibility	none
Reasoning	consequentialist
Policy	none
Emotion	mixed
Coded at	2026-04-27T06:24:53.388235

Raw LLM Response

[
{"id":"ytr_UgxBSPLgIZgoxW75T_54AaABAg.AT5zqZXavDIAT7Flt59JVm","responsibility":"ai_itself","reasoning":"consequentialist","policy":"unclear","emotion":"mixed"},
{"id":"ytr_UgzqAFwsAv_KFYodnsZ4AaABAg.AT5zYObdugsAT6N8mi1y-k","responsibility":"none","reasoning":"consequentialist","policy":"none","emotion":"fear"},
{"id":"ytr_UgypFt-geNkrVXiCZKZ4AaABAg.ACqSJ-64r_IADGqQZSIyda","responsibility":"developer","reasoning":"deontological","policy":"none","emotion":"outrage"},
{"id":"ytr_UgypFt-geNkrVXiCZKZ4AaABAg.ACqSJ-64r_IADHiqhf0--s","responsibility":"developer","reasoning":"deontological","policy":"none","emotion":"mixed"},
{"id":"ytr_UgwmijS6ORinBNiU97l4AaABAg.ACPn9QBJA7kACXUpFJQjzw","responsibility":"none","reasoning":"mixed","policy":"none","emotion":"approval"},
{"id":"ytr_UgwmijS6ORinBNiU97l4AaABAg.ACPn9QBJA7kACXkGqvjRfO","responsibility":"none","reasoning":"consequentialist","policy":"none","emotion":"mixed"},
{"id":"ytr_UgwmijS6ORinBNiU97l4AaABAg.ACPn9QBJA7kACYsef1X_0Z","responsibility":"developer","reasoning":"consequentialist","policy":"ban","emotion":"fear"},
{"id":"ytr_Ugw5RY3Jgi5ZKd0rJTh4AaABAg.ACPiaRNcYV4AD6EVM4dvv8","responsibility":"developer","reasoning":"consequentialist","policy":"regulate","emotion":"mixed"},
{"id":"ytr_UgxPWTX-Ney8A1b-AYh4AaABAg.ACOTh2V31EVACQDPHy_sHM","responsibility":"none","reasoning":"mixed","policy":"none","emotion":"indifference"},
{"id":"ytr_UgxPWTX-Ney8A1b-AYh4AaABAg.ACOTh2V31EVACRLOiBBt8k","responsibility":"none","reasoning":"mixed","policy":"none","emotion":"outrage"}
]

Raw LLM Responses