Pretty much the title. I certainly believe Trump, Maga, the Military, and the Police will, sooner or later, probably sooner, get around to at least attempting to deanonymize and round up online antifascists and leftists and imprison them. How organized and effective that attempt is I am less sure of.
To be very transparent, this is something I’m pretty sure I’d be on the hook for. I have a long log of anti trump, antifascist, left sentiments, and am 75% sure I’ll be disappeared at some point in the next 4 years as I have no plans of shutting up. The only reason I’m not 100% sure is because of how expensive it would be. But hey, maybe it’s less expensive than potentially losing power? So I don’t know.
Never? Not likely? Maybe? Very? Extremely? Definitely?
Thoughts?
Ah yes, as always, the ever-knowing magical AI. Because all the secrets of the universe can be aggregated using just few multiline prompts to the ChatGPT.
You’re distrust of AI hype is fine, but you’re missing the point. OP said AI makes it simple to compile dossiers on everyone, meaning it’s now far less labor intensive to take all of the data being gathered by SIGINT and turn it into reports. The amount of labor required to build 10M dossiers on mostly impotent randos makes it completely unfeasible, but with generative AI being able to quickly summarize a dataset, suddenly we can have shitty, somewhat lossy dossiers on every moron shitposter.
I’m pretty sure the capabilities were there before LLMs and the LLMs don’t add that much value relative to it’s cost.
The capabilities to generate summaries? Yeah, that’s probably true, I knew of AI that could summarize data into narrative years before the current LLM hype. I don’t know the underlying tech behind those things I saw though. They could have been early LLMs, they could have been some other neural network. But it was definitely a machine learning solution.
The point that machine learning can reduce the effort to produce dossiers to the point of making millions of dossiers feasible still stands though.
It’s not even an AI thing; look at how widespread domestic surveillance was back in the 60s. In the years since it’s only gotten easier to record messages and calls, match those records back to addresses or GPS-given locations, catch people on video, etc.
Not chatGPT. But AIs are good at one single thing and that is pattern recognition. And there is a lot of data to train and use those AIs on. Decades of data from your PCs, phones and tablets is on NSA servers. And yes, the NSA did shit like recording you through your notebook camera or recording your phone calls with your first girlfriend.
AI will probably never replace human workers/engineers/artists/…, but sorting through your online history is exactly what AIs excel at.
Also Dragnet policies are used by the police for a long time now. This is pretty much just that. Imo the main question is more if they want to burn that data on this task, because once they do this some people will adjust and it will get harder to get new data.