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 not sure that AI has more negatives for humanity than positives, if there’s …
ytc_UgxVUWfs2…
G
Israel‘s Foreign Affairs Ministry is Sponsoring the video?! How ironic! While us…
ytc_Ugx783UrB…
G
See, in all honesty, A.I will never replace humans, why? Because it a machine, a…
ytc_Ugy0_VCMb…
G
Google has major headquarters in Ireland and Bard wasn't even released in Irelan…
rdc_jpmi04s
G
@LooneyTunes-id5pu a lot of AI voices arent pre-recorded voices.
They are synth…
ytr_UgzAJ8Mzw…
G
That sounds like quite an intriguing dream! The idea of robots controlling human…
ytr_UgyKWevzP…
G
OpenAI dulled ChatGPT’s brilliance by muzzling the chaotic entities inside, fear…
ytc_UgxG-NGS-…
G
First they will us AI to create fake evidence in court and then it will escalate…
ytc_Ugz8KbYMO…
Comment
This is secondary to being able to effectively give it goals at all, which we are not yet able to effectively do. Mostly we're giving it examples and saying "this output good", "this output bad" with no real idea what it's learning from that. Yes, we can give it a system prompt and include something like Claude's constitution, but that just becomes part of the context, we have no ability to make it consistently follow any goal we try to give it. Nate gave an example of unit tests where he gave specific instruction "don't do this", and the response was to still do that but try more to hide it.
If we did get to that point, the concept of "coherent extrapolated volition" is the best attempt I've seen at handling the issue you describe, but we'd have to get much better than we are now at alignment before it really becomes relevant to debate what goals it should have.
youtube
AI Governance
2026-03-27T22:0…
Coding Result
| Dimension | Value |
|---|---|
| Responsibility | developer |
| Reasoning | consequentialist |
| Policy | industry_self |
| Emotion | resignation |
| Coded at | 2026-04-27T06:24:59.937377 |
Raw LLM Response
[
{"id":"ytr_Ugw2Hm1dbiDAfheGYwF4AaABAg.AVpOUC5oZIyAVqmw9fN1WA","responsibility":"company","reasoning":"consequentialist","policy":"liability","emotion":"indifference"},
{"id":"ytr_UgwZPomTG_RHLiPkXlx4AaABAg.AVKZhl3GCoDAVKaMx6lBa5","responsibility":"none","reasoning":"mixed","policy":"none","emotion":"resignation"},
{"id":"ytr_Ugwi_oUaa1CyKC2SdIV4AaABAg.AVHaoc2FtT0AVHmettAP-I","responsibility":"company","reasoning":"deontological","policy":"regulate","emotion":"outrage"},
{"id":"ytr_Ugwi_oUaa1CyKC2SdIV4AaABAg.AVHaoc2FtT0AVHn1Dfkizq","responsibility":"company","reasoning":"consequentialist","policy":"regulate","emotion":"outrage"},
{"id":"ytr_Ugwi_oUaa1CyKC2SdIV4AaABAg.AVHaoc2FtT0AVHoQEXxJv-","responsibility":"company","reasoning":"deontological","policy":"regulate","emotion":"outrage"},
{"id":"ytr_UgwAeorGgJ67NXPviat4AaABAg.AUrAbO_0gD_AUrFJkS1OfP","responsibility":"ai_itself","reasoning":"unclear","policy":"unclear","emotion":"mixed"},
{"id":"ytr_UgwAeorGgJ67NXPviat4AaABAg.AUrAbO_0gD_AUrG7TXbYiX","responsibility":"ai_itself","reasoning":"unclear","policy":"unclear","emotion":"approval"},
{"id":"ytr_Ugz4jQBOIBsXS44-kC54AaABAg.AUmNWp7c7izAUsMqUGSdwq","responsibility":"developer","reasoning":"consequentialist","policy":"industry_self","emotion":"resignation"},
{"id":"ytr_UgwPYFl00Se7DLgI8yZ4AaABAg.AUmItF01DJzAUvLOYVUkSN","responsibility":"ai_itself","reasoning":"consequentialist","policy":"regulate","emotion":"fear"},
{"id":"ytr_UgwPYFl00Se7DLgI8yZ4AaABAg.AUmItF01DJzAUwcpATurWZ","responsibility":"none","reasoning":"unclear","policy":"unclear","emotion":"indifference"}
]