Great story, feels like fMRI type studies where they ask questions and parts of the brain light up.
Very similar indeed to the fMRI studies; which gave me minority report vibes, where you could be picked up and charged for “precrimes” by just having a thought… and much like that particular technology, I would like to see significant amount of oversight on having the ability to “dial in” LLM model responses.
It would be quite tempting for someone to (or pay for) a shift in political sentiment within the AI responses…
As such, “dialling in” AI responses to shift the overton window has been added to my list of concerns 
2 Likes
Bro - the US already has the table tilted pretty heavy towards them, so this might help rebalance that (as well as all of the bad stuff you identify).