[nisegami]

I personally think that frontier LLMs are already at the point where I trust them to be a better judge of human behavior than humans overall.

[Schiendelman]

Absolutely. But they still hallucinate, I wouldn't want them to take action themselves, just flag things for human review.

[baranul]

Even in that context, to just be constantly flagging human actions, can be argued as reflective of a dystopian future. Always being monitored, and every viewable action recorded, categorized, and flagged.