Hacker News Reader

Comment by chiefalchemist | Hacker News Reader

> The system knows everything about [the terrorist]: where he went, who his friends are, who his family is, what keeps him busy, what he said and what he published. Using artificial intelligence, the system analyzes behavior, predicts risks, raises alerts.

Where does "the terrorist" end and me, you and anyone else just minding our own business get inserted instead? And let's say it's not even the gov doing this but some private company with public data, what's to stop the gov from buying "reports" from that company. 100% legal. That is, no rights being violated, etc.

Anyone who says, "I have nothing to hide" is a fool, at best.

View on HN

The terrorist ends wherever a threat to the state's power exists, it's been shown quite well that they don't care if you're domestic or not. This, IMO, is why "self-radicalized" and "domestic terrorist" were injected into our vernacular, to normalize and justify the need to surveil the general public.

The thing is, and like I mentioned in a post awhile back: technically competent actors, the ones bound to cause the most harm, would absolutely be using a bespoke method of covert communication. There's really little value, IMO, to the countrywide dragnet outside of sentiment analysis and control.

The military complex wove itself early into the tech industry in ways that they could intentionally side-step laws meant to keep such public/private collusion from happening[0]. The impetus for the founding of the collaboration was a report on the importance in controlling perception in future wars.

We saw the same strategy deployed directly against the American people during the election "fortification" where DHS and social media colluded to control perception with little regard for truth[1].

[0] https://medium.com/insurge-intelligence/how-the-cia-made-goo...

[1] https://homeland.house.gov/2023/11/06/chairmen-green-bishop-...

> The thing is, and like I mentioned in a post awhile back: technically competent actors, the ones bound to cause the most harm, would absolutely be using a bespoke method of covert communication. There's really little value, IMO, to the countrywide dragnet outside of sentiment analysis and control.

Agreed. And yet their persistence to surveil continue to expand. They're not trying to watch those who are sure to be hiding. They're watching the rest under the guise of "We gotta get those terrorists."

I don't support ubiquitous spying at all but are you hanging out with known ISIS members or members of White Nationalist Militias regularly? Because I'm pretty sure that's where the line begins.

There's a lot of grey here. What does "hanging out" mean? If my weird uncle is (unknown to me) in ISIS does spending thanksgiving with him count as "hanging out"? ISIS is at least pretty specific, but what counts as a White Nationalist Militia? Both of these can be redefined to capture more and more of the population if desired.

This is what we have attorneys and judges for. And no, obviously Thanksgiving isn't "hanging out". But going to the same weekly meeting and practicing lynching minorities? Yeah, that's a little more than just mashed potatoes and gravy, isn't it. (These contrived "whatabout" gotchas are exhausting. It is abundantly obvious who is and who is not involved with white nationalist militias.)

It's really the court of public opinion that has the greatest risk of harm at the day to day level. A non-poc going to the gun range and then posting on social media could cause a "White Nationalist Militia" label to get attached by a jilted coworker and then go viral. That can cause serious harm.

Maybe gun clubs should implement DEI programs. Then those pictures would have some colorful people in the background.

Yes, all the legal arguments presented before the FISA court by the lawyers working on behalf of those targeted have been really interesting reads!

There aren't legal arguments by the targets in law enforcement search (either physical or wiretap) warrant cases either, that mainly only happens if (as does not always happen) the product of the search is used in criminal prosecution later.

None of those things seem particularly obvious to me.

I think they're just saying it's a slippery slope. It starts out with good intentions we all agree on, but then continues to slide and more and more of our freedoms erode as they crank up the boiling pot ever so slowly.

If you don't know, who they are and just happen to serve a 'bad guy' ( hate that term ) a burger, should you be in the crosshairs? Because this is where this is heading. If you think I am overselling it, remember that police in US can ask for all users from specific location.

To your point, if there is indeed a line, it need to be clearly articulated so that the rules of the game clear.

You've heard of guilt by association, well guilt by co-location isn't that far off. Along the same lines, the rationale is going to be, "We need to track everyone so we can be sure to see all relationships and connections, and connections to connections, and so on.

Try to draw the line wherever you want, but they're going to step over it, and never look back.

Maybe that's where the line starts, but does it stay there? As an example, look at how the US's anti-communist fervor led to things like COINTELPRO: https://en.wikipedia.org/wiki/COINTELPRO

Especially today, I think we have to look at every power we might give to government and ask, "What happens if the worst people get access to this?" Because they're certainly going to try.

Did we give it, or was it taken? Yeah, maybe we consented to The Patriot Act. But when it was renewed, it was *expanded*. Too late now.

> I'm pretty sure that's where the line begins

Based on what?

From my perspective, the easiest way to design such a system would be to create entries for every 'actor' in the system, feed in as much data as you can get your hands on, and then let the weights sort themselves out. So for example, if you're hanging out with ISIS members obviously your weights would be higher, but even if you're a server at Applebees you'd still be in the system somewhere.

Doing it the other way necessitates some kind of bright-line division, and any such boundary, once defined, becomes susceptible to exploitation. e.g. I won't hang out with the White Nationalist Militia because that puts me "into the system", but I can hang out with insert radical right-wing group where I can talk to 80% of the same people without being flagged. In practice, I imagine that the gradient of extremism is rather gradual and with blurred boundaries.

As another poster mentioned this is literally why we have courts. There is a clear line for obtaining a search warrant for example. Precedents exist.

Are you familiar with FISC? I'd say go familiarize yourself with its case law, but you can't, because it's secret. And it authorizes methods much more powerful and invasive than a simple search warrant. Precedent exists, but nobody outside the national security state actually knows what it is.

You're maybe proposing another line where no spying is legal at all and we should just submit ourselves to the whims of terrorists or other lunatics? Surely there is actually a line where there are tradeoffs between security and privacy and its probably not 0% security and 100% privacy.

Perhaps you think all FISA rulings should be public and any sufficiently savvy malicious actors can just read them to know exactly how to avoid suspicion?

Everything the FISA process overseen by the Foreign Intelligence Surveillance Court and the Foreign Intelligence Surveillance Court of Review can authorize (and much more invasive means, contrary to your claim) can be authorized by regular search warrant.

The FISC process is used when the purpose is foreign intelligence rather than domestic law enforcement, and it exists because prior to that there was no limit on the covered activity when it was done for that purpose.

> Precedent exists, but nobody outside the national security state actually knows what it is.

Well, some of it.

https://www.fisc.uscourts.gov/public-filings

https://law.justia.com/cases/federal/appellate-courts/fiscr/

I think, in a sense, that part is already over. Entities that encompass both ends of the spectrum exist and any remaining gaps are filled by public/private partnership ( and hailed as a great thing just about anywhere ).

The scenario of a private corporate entity wielding that power has already come to fruition if you look at what Google or Facebook has available on its users.

I think that is the main reason why I am not as.. restrictive on use of LLMs and AI, because I see it as a form evening out the playing field at least a little bit.

I think one of the things that scares me most about API-accessed LLMs is how powerful they are as data collection tools in their own right. OpenAI, for example, recently updated their terms of use to be more vague about how they work with the gov and I have no doubts that giving the NSA access to conversational feeds is absolutely a requirement to their continued operation as an entity, a la lavabit.

In fact, part of me thinks that the Sam/Ilya drama and sam's god complex are at least partially rooted in this, alleged, collaboration.

Imagine the questions you could pose to a GPT trained on all the conversations had with users that's been enriched with their biographical data. These conversations are often intimate and curiosity driven in a way that seeking the truth could easily be framed as self-radicalization.

That API is one the main reasons companies are not as keen on jumping on the bandwagon. They don't want to have OpenAI to have access to their corporate data. But then, there are options of running models locally..

I think your concerns are valid.

Reply on news.ycombinator.com