Raw LLM — Corpus Dashboard

Look up by comment ID

Random samples — click to inspect

G Are we really living in a timeline where Palpatine invented AI and later wanted … ytc_UgwALP8Rq… G MAD is more or less a concept humans have decided on to feel better about wieldi… rdc_o7c5v4k G Humans are just cacoons for A.I, oir story doesn't end with it because we will l… ytc_Ugx2MboH5… G At first I thought this was kind of real but as I see it more it seems you are p… ytc_UgyAacWZd… G @MisterMyagi: Thank you, appreciate you! Your comment made me laugh harder than … ytr_UgwSbZsIn… G That’s bullshit about using and wasting drinking water for datacenter cooling, … ytc_UgzLPMIlA… G A.I is a lie and a scam. It only benefits companies who want to replace human do… ytc_UgywwvKGg… G Too late for that. Aren't you keeping up with the news? Optimus robots are antic… ytr_UgzoXd4BZ…

Comment

To suggest this case and others like it are evidence that LLM based AI agents will not imminently and competently displace human lawyers is naive. A single LLM like GPT-3 or GPT-4 is like a genius with no capacity for reflection. Speaking the first thing it thinks with no review. If an accurate answer is not forthcoming (e.g. how to win an unwinnable case), it naturally makes something up because that's what is most consistent with being helpful (being helpful is it's prime directive). In this case, GPT-3 didn't fail. It did a superb job of exactly what it was set up to do: predict what a lawyer would most likely say given the absolute premise that the lawyer has something helpful to say. This is already a very well understood and solvable problem. Advanced techniques such as "chain of thought", and simply incorporating accuracy into the reward function, are already yielding promising results. None of these techniques were deployed in the case in question, so failure was a likely result.

youtube AI Responsibility 2023-09-13T21:3…

Coding Result

Dimension	Value
Responsibility	ai_itself
Reasoning	consequentialist
Policy	none
Emotion	indifference
Coded at	2026-04-27T06:26:44.938723

Raw LLM Response

[
{"id":"ytc_UgzDGZfGItK9LX2OAMR4AaABAg","responsibility":"ai_itself","reasoning":"consequentialist","policy":"none","emotion":"indifference"},
{"id":"ytc_UgzdYPkIAaHaRkeQuil4AaABAg","responsibility":"none","reasoning":"consequentialist","policy":"none","emotion":"approval"},
{"id":"ytc_UgzYx8QkEw5qEIlVa_V4AaABAg","responsibility":"user","reasoning":"mixed","policy":"industry_self","emotion":"mixed"},
{"id":"ytc_UgyStPwsGTg_T6SQsMV4AaABAg","responsibility":"user","reasoning":"deontological","policy":"none","emotion":"outrage"},
{"id":"ytc_UgxowcuLHbRZgbj5O7d4AaABAg","responsibility":"user","reasoning":"consequentialist","policy":"none","emotion":"outrage"},
{"id":"ytc_Ugxts-i6mZS9sahdjeR4AaABAg","responsibility":"none","reasoning":"consequentialist","policy":"none","emotion":"indifference"},
{"id":"ytc_UgyBK4e_MpA_S778Hlp4AaABAg","responsibility":"none","reasoning":"mixed","policy":"none","emotion":"mixed"},
{"id":"ytc_UgyE0HauJPrrOzypmcd4AaABAg","responsibility":"user","reasoning":"consequentialist","policy":"none","emotion":"outrage"},
{"id":"ytc_UgxiWicCvPsoko-JQc14AaABAg","responsibility":"none","reasoning":"unclear","policy":"none","emotion":"fear"},
{"id":"ytc_UgzPyi3Axuars712SjV4AaABAg","responsibility":"ai_itself","reasoning":"deontological","policy":"none","emotion":"indifference"}
]

Raw LLM Responses