Yeah, I'm not a fan of these things. If they were just ALPRs, I could probably give them a bit of slack -if they tightened up their security-, but all the other stuff they do, makes them pretty much untenable.
However:
> This makes AI powered cameras like Flock's distinct from traditional surveillance or traffic cams, which require someone to manually look over footage in order to find a specific vehicle or individual.
Is a bit misleading. These days, anyone can give an LLM footage from any source, and get this kind of information.
What LLM can I get and feed hundreds of hours of video into that will give me the position of a specific vehicle alongside when that happened?
An LLM isn’t going to help you here, but basic Computer Vision and a SQL database has been a solution _if you have the cameras_. I wrote a license plate reader as a university project using OpenCV almost 20 years ago.
If you were able to write one 20 years ago, I dare say an LLM could whip one up super-fast. Or just search the internet and tell you where to find one.
One of the risks of LLMs is that a lot of tasks go from "an expert could do this easily given a few weeks" to "anyone who thinks to ask an LLM can do this easily and get results the same day"
> One of the risks of LLMs is that a lot of tasks go from "an expert could do this easily given a few weeks" to "anyone who thinks to ask an LLM can do this easily and get results the same day"
By that logic, all problems are solved with LLMs, though.
Honestly, there's plenty of experts worried about exactly that. But if you'd prefer, feel free to mentally scope my claim to the domain of "things we already know LLMS can do, like copying previously solved software problems"
Dropbox stock is trading at 50% of its initial price 5y ago when it went public, maybe the public markets also don't understand the difference between rsync and Dropbox.
I honestly can't tell if this is ragebait or you believe this.
My friend, if you have a database of license plates extracted from single images taken by multiple cameras, YOU ARE TRACKING UNIQUE VEHICLES ACROSS A REGION.
Terabytes of data don't matter because you don't need to search terabytes, you need to search a few MB of text data. You don't even have to store the original video.
You’re moving the goalposts. The original point of this thread is that Flock AI technology is hardly needed to efficiently search traditional video footage for license plates.
> Assuming they have access to terabytes of regional video surveillance but don't have their own compute besides what the LLM will buy for them
Amazing how you can move the goalposts to make things impossible, isn't it? Where in the world did "without compute" come from? Are they not even allowed a decent desktop computer?
Any competent llm would write a script using opencv to extract the license plates.
I did this with Gemini 3, mostly for fun and to test it's capabilities. Teslausb records all dash cam videos and auto syncs it to my nas when in wifi range. Yolo and opencv extracts and does ocr on any defected license plate, and puts it all on a map, along with trip information. Not particularly useful or interesting, and not something I would have done pre-llms, but the difficulty was basically writing a one paragraph prompt and using some free tokens
I think the key word is on "any one". Personally, I know how do to the work, I just didn't want to because it provided minimum value to me. But I didn't actually need to know how to do it, because the llm produced perfect results for me without needing to prompt it with any specifics
But what you have is a license plate reader. They've existed and are a solved(mostly) problem since the 80s. Flock isn't an ANPR, it's a network of cameras that has that as a feature.
An LLM isn't replacing an interconnected camera network, it's replacing a script that can be replicated from a 6 year old blog post!
You made this same incorrect claim in another comment. You’re generalizing beyond what others are claiming, without any basis. (This is the logical fallacy known as hasty generalization.)
What makes Flock bizarre is that it's a private business, and this is precisely how police departments are getting around a lot of traditional gates and checks on this sort of thing.
Police setting up a 1984 monitoring system throughout your city, tracking every car, person, activity -- yields lots of questions, oversight, concerns, debate, challenges, etc.
Some private business doing the same, and then letting the same police use it at will as a paying customer -- yay, all of the invasive monitoring with none of the oversight.
And of course, it's compounded by being pooled. Like RealPage, ALPR services like Flock, Axon Fleet Hub, and Motorola Vigilant VehicleManager offer data laundering so that organizations that shouldn't be talking can communicate.
The real joke is that the 1984 system was already in place long before flock. We all carry phones. Either from tower records ot google advert tags, private industry already tracked our every movement. Licence plate readers? Why bother when every car now has bluetooth enabled and so is constantly prodcasting its LAP id. (Some cities have tracked cars this way for over a decade, mostly for traffic management.)
Police had to at least play 'parallel reconstruction' and pretend they weren't using Stingrays though. And we all knew it and nodded along, and the lines moved, and the rules blurred a little more, and now we are at the next phase of the erasure of civil liberties.
That's actually the one thing that does make sense: police has always wanted to be able to do this, but they legally can't. But they can reward a private company willing to do it for them, so that they can "ask for the data" without ever breaking the law.
The police can, in fact, operate cameras in public spaces and they have done so for decades. ALPRs have been widely deployed since the 1990s.
I'm frequent surprised by how many people think that privacy laws block the police from recording their activities in public. For whatever reason, Flock is getting a lot of press, but this is hardly a new field.
ANPR has not been widely deployed since the 1990's in the US and the US court system has consistently held that the degree to which a technology automates monitoring, searching, etc is very relevant to whether it violates people's reasonable expectations of privacy for very obvious reasons.
> However, ANPR did not become widely used until new developments in cheaper and easier to use software were pioneered during the 1990s. The collection of ANPR data for future use (i.e., in solving then-unidentified crimes) was documented in the early 2000s.
Nah. Fascism only tolerates one power, that being itself. It can emerge from either the state or corporate side, and necessarily subsumes or destroys the other, just as it subsumes or destroys unions, families, friend networks, communications, and anything else that can establish power. That doesn't mean the merger of two of them is the defining feature.
Corporations generally tend to only tolerate the state to the extent that guns or courts mandate that they must. How many billions of corporate dollars have gone to fund campaigns to deregulate, to skirt authority, to do whatever is necessary to make sure profits go up?
No. Power can be delegated. I'd say capitalism puts ultimate power on the government who delegates power to holders of capital. (If that stops, it's not capitalism any more)
There's a very important difference between "anyone could walk through my door and steal my stuff" and "this person walked in my door and stole my stuff".
But not so much between "this person walked in my door and stole my stuff" and "I left the door to my house open and then I put my stuff in the doorway."
Personally I don’t mind having Flock cameras or other surveillance to make it easier to stop crime. Right now in many cities there is simply no consequence for various crimes - especially property crimes. But I want these tools to be protected - requiring imminent danger or warrants for access. I don’t think the answer is to get rid of them entirely.
That's interesting, because the ALPR part of Flock is what caused all the problems here; the rest of it, of characterizing vehicles with attributes beyond just plates, wasn't really problematic at all.
Characterizing vehicles based on stickers (which easily can indicate political leanings) is absolutely problematic. It's just not considered problematic by conservatives, given police in the US are overwhelmingly conservative.
Look, I agree that's problematic conceptually, but it's absolutely not what's happening. ALPRs don't select cars on spec, like, "this car looks out of place here". Nobody has time for that. They're matching specific descriptions of cars to incidents, like, "this vehicle has been present at the site of 5 previous package thefts".
Here's another way to look at this. Municipalities are the primary operators of ALPR cameras. Any municipality that would scan bumper stickers looking for Trump opponents is not going to be receptive to any appeals for regulation.
One problem with this whole debate is that people are coming to it with movie plot concerns rather than understanding what's actually happening with them. That wouldn't be a big deal if this was a slam dunk public policy case, but it isn't: there is broad bipartisan support for these devices.
There are deeply problematic things happening just with license plate pings!
They're matching specific descriptions of cars to incidents, like, "this vehicle has been present at the site of 5 previous package thefts".
You're hand-waving a hell of a lot of things away and you expect that everyone knows what you're talking about. Please stop doing that.
- Who is "They"?
- Why do you say "nobody has time for that"? What is "that"?
- Why are you dismissing genuine concerns through unhelpful language like, "coming to it with movie plot concerns".
- Why wouldn't "that" be a big deal? What is "that"?!
- What are the deeply problematic things?
- "They're matching specific descriptions of cars to incidents" -- no they're not. Just looking at Bloomingdale's audit logs, there are 13k examples of searches done for the simple reason, "suspicious".
- Why does municipalities being the primary operators matter?
Asking from a place of genuine confusion by how you think about these things.
"They" are public bodies operating ALPR devices; in the main, municipal police forces, though obviously other public bodies (like the Illinois State Police) operate them as well.
The antecedent of the first "that" was "the use of cameras to detect cars with politically disfavored bumper stickers".
I am, yes, dismissing the concern that ALPRs are being used to detect cars with politically unfavorable bumper stickers. I think that if advocates for ending our Flock contract had come to the board table with that concern, rather than the quality of Illinois LEADS, we'd still have the cameras up.
The antecedent of the second "that" is "organizing around easily dismissible movie-plot concerns, like that municipal police are going to dragnet for people with anti-police bumper stickers". Unwinding the sentence, the "big deal" is, as I just said, that centering implausible risks takes real risks out of focus, and gives ammunition for advocates of the cameras --- of which there are a great many --- to push back on efforts to get the cameras down.
I spelled the "deeply problematic" things out elsewhere on the thread.
Feel free to tell me more about what Bloomingdale was doing with their cameras. With no detail, I'm inclined to believe the force simply didn't give a shit about the description field in the search request, because no serious, rigorous effort was made to regulate ALPRs in Bloomingdale, and so there isn't much signal in the logs.
My common experience is that people dismiss these risks without evidence, while I've seen plenty of stories of such things happening in this context and in others. Possibly those stories add up to a lot of anecdotes, but these aren't arguments based on reason or evidence:
> easily dismissible movie-plot concerns
> implausible risks
Those are just words. People who use them, IME, imply their conclusions are already well-established. But they never are. Where is the evidence and argument for these claims?
Please actually just look at any audit log and just search yourself. If you think there's no signal, then you clearly haven't looked. If you're going to continue to be lazy in your analysis, then ask a damn LLM. There are 528 agencies who used "suspicious". This is not a bloomingdale problem; it's much larger. Just fucking look, man.
select count(*),org_name from flock_bloomingdale where reason = 'suspicious' group by org_name order by count desc;
count | org_name
-------+--------------------------------------------------
2678 | Skokie IL PD
828 | Joliet IL PD
678 | Houston TX PD
391 | Fayette County IL SO
309 | Chicago IL PD
256 | Katy TX PD
245 | Itasca IL PD
244 | Steger IL PD
229 | Athens-Clarke County GA PD
215 | Lucas County OH SO
209 | Oak Lawn IL PD
208 | Westmont IL PD
199 | La Salle County IL PD - OLD
194 | Zion IL PD
191 | La Grange Park IL PD
174 | Kenosha County WI SO
173 | Champaign County IL SO
170 | Roselle IL PD
160 | Lake Villa IL PD
152 | Bradley IL PD
152 | Madison County IN SO
143 | LaSalle Co. IL SO - New
135 | Flossmoor IL PD
132 | Sauk Village IL PD
116 | Oak Brook IL PD
106 | Crete IL PD
104 | Villa Park IL PD
101 | Darien IL PD
97 | Cicero IL PD
94 | Wilmington IL PD
89 | Rockford IL PD
80 | Lake County IL SO
80 | Dolton IL PD
79 | Texas Department of Public Safety
76 | Will County IL SO
75 | Naperville IL PD
72 | Minooka IL PD
68 | Hillside IL PD
63 | Carpentersville IL PD
55 | Kent County MI SO
55 | Zanesville OH PD
54 | Winnebago County IL SO
51 | Logan County NE SO
46 | Romeoville IL PD
46 | Menomonee Falls WI PD
46 | Homewood IL PD
44 | Burnham IL PD
44 | Baldwin County GA SO
43 | Venice FL PD
39 | Elmwood Park IL PD
37 | DuPage County IL SO
36 | Greensboro NC PD
34 | Lowndes County GA SO
34 | Henry County GA PD
34 | Tinley Park IL PD
You're going to need to tell me why I'm supposed to be freaked out by the word "suspicious". Further, it is not my claim that (1) ALPRs are not problematic or (2) that they don't direly need more regulation; that's exactly the work I was doing in Oak Park before we cancelled our contract.
hmm, ok. kind of odd to put yourself out there for so many years, then be willing to die on this issue, given the political state of things. but, you be you
Edit: I can't respond to your reply, but my original question was why is it so important to you, and you didn't answer. But, I guess I'll go back through your comments to find out. Us 'passive observers' are just trying to figure out who's on which team. Feel free to ignore us. No offense intended
"Weird that you sound so defensive, given you're willing to die on this issue and ruin your reputation". Do you hear yourself?
What's especially funny about this is that you haven't taken the time to even figure out who I am or where I'm coming from on this issue. It's pure uncut schoolyard behavior. Maybe reflect a bit!
Nobody's reading this thread but us at this point.
However:
> This makes AI powered cameras like Flock's distinct from traditional surveillance or traffic cams, which require someone to manually look over footage in order to find a specific vehicle or individual.
Is a bit misleading. These days, anyone can give an LLM footage from any source, and get this kind of information.