As far as I know, LLMs are not that clever yet, and it would require a lot of work to automate tracking of so many targets. But a dedicated person tracking one user can see these. Unknowingly, we leave a lot of cues to know who we are. Not only patterns, but exact word-markers, like calling something by a regional-accepted name. Like how my english teachers insisted London’s metro is called Tube.
PoisonedPrisonPanda@discuss.tchncs.de 11 months ago
Yeah I am not interested in criminal behaviour or personal threats. i know a human would be capable to extract much information if on purpose.
I am more interested in like temperature mapping of my text. From a statistical point of view If my patterns are behavioral and forecastable?
andrew_bidlaw@sh.itjust.works 11 months ago
It’s mainly keywords at that point. This process is sure to have steps. To step into a suspected category, to be elevated into those who are to look closer, you should ring some alerts.