Raw LLM Responses
Inspect the exact model output for any coded comment.
Look up by comment ID
Random samples — click to inspect
G
Next would be to run GPT4 against a specialist several times to see if the diffe…
ytc_UgxgB-gEi…
G
Artists should upload shitty art on the internet so the AI would also eat that d…
ytc_UgxQH4k3W…
G
Bro, I'm already tired of people trying to "cancel" people who make AI drawings,…
ytc_Ugysrhz1e…
G
Where is this money coming from? The AI companies and billionaire class sure as…
ytc_Ugz5DXFQR…
G
5 years ago, the concept of chatgpt would have been completely alien and imposib…
ytr_UgzqRAIOy…
G
That's totally very possible giving right freedom machine sentient can doing eve…
ytc_Ugz-B0Yzd…
G
Just a whole lot of jealousy that you couldn't make it as a professional artist …
ytc_UgzGKZW01…
G
Relying on ChatGPT to do your thinking for you can lead to trouble. It's importa…
ytc_UgzVHOXvI…
Comment
Eliezer's take on the 'paperclip maximizer' argument doesn't seem particularly applicable to current LLM architectures. When I ask ChatGPT for an answer, it neither gets stuck in an infinite loop nor produces endless responses in an attempt to 'maximize' its objective. Working with agents also involves setting constraints: we can specify a finite number of actions the model should run, and there's a system of permissions to accept or deny subroutine actions. It's unclear why Mr. Wolfram didn't tie this argument to known, practical AI procedures.
Also, if AGI truly achieves human-level general intelligence, it would presumably possess practical judgment capabilities. ChatGPT, for instance, provides finite responses rather than infinite outputs, and an AGI would theoretically have even more refined judgment. Just as adults have better risk assessment skills than children, an AGI should theoretically evaluate actions within realistic limits rather than pursuing infinite maximization of a single goal.
youtube
AI Governance
2024-11-13T16:0…
Coding Result
| Dimension | Value |
|---|---|
| Responsibility | developer |
| Reasoning | consequentialist |
| Policy | none |
| Emotion | indifference |
| Coded at | 2026-04-27T06:24:53.388235 |
Raw LLM Response
[
{"id":"ytc_UgwfYHnRIec_UjaORrV4AaABAg","responsibility":"none","reasoning":"unclear","policy":"none","emotion":"approval"},
{"id":"ytc_UgycnzNreGpB3a7a5Hp4AaABAg","responsibility":"none","reasoning":"mixed","policy":"none","emotion":"indifference"},
{"id":"ytc_Ugzd-ma0ujZAb5HhHFp4AaABAg","responsibility":"none","reasoning":"consequentialist","policy":"none","emotion":"mixed"},
{"id":"ytc_UgzsZtPkhMQCcCOmHgB4AaABAg","responsibility":"developer","reasoning":"consequentialist","policy":"none","emotion":"indifference"},
{"id":"ytc_UgxYn9JXLlg20G_a09d4AaABAg","responsibility":"none","reasoning":"consequentialist","policy":"none","emotion":"outrage"},
{"id":"ytc_Ugz2_DwgYk7tALNnvm54AaABAg","responsibility":"ai_itself","reasoning":"consequentialist","policy":"none","emotion":"fear"},
{"id":"ytc_Ugwad4p8PY-nWvnjzPN4AaABAg","responsibility":"none","reasoning":"consequentialist","policy":"none","emotion":"outrage"},
{"id":"ytc_Ugx0w3H6RV1sNvUp1ZV4AaABAg","responsibility":"none","reasoning":"mixed","policy":"none","emotion":"mixed"},
{"id":"ytc_UgyR6_fTp_kjrcdO_SV4AaABAg","responsibility":"none","reasoning":"unclear","policy":"none","emotion":"indifference"},
{"id":"ytc_UgwxlrHOJKfspbgJ1TZ4AaABAg","responsibility":"none","reasoning":"unclear","policy":"none","emotion":"mixed"}
]