Hacker Newsnew | past | comments | ask | show | jobs | submit | mlmonkey's commentslogin

This wouldn't be a story if the cops did not put the wrong license plate in the system. How is it Flock's fault? Flock is just doing what it is being asked to do!

Let me put in simple terms: Flock flags license plates that are given to it. Someone, somewhere says, license plate "ABCD1234" has a warrant out. And guess what, if Flock sees that plate, it _will_ flag it each. and. every. time!

Tomorrow, say an "Amber Alert" is issued for a pink Ford Taurus with plate "PINKLADY" (when in fact it was a red Taurus with the plate "MADLAD"). Don't you think anyone driving around in a pink Ford Taurus with that plate should be pulled over?


How are all these dead baby seals Flock's fault? They simply released the Auto Baby Seal Clubber 9000 on beaches that have baby seals. It's the people that keep submitting "club baby seals" to the system that are the problem.

What I want to know is who is manufacturing these police cars that let these cops travel to execute unlawful warrants. "Oh, but it's not our fault. We just built due-process-violation machines. It's the police who are driving them to citizen's locations and violating due process." Come on.

Once? Maybe. And then the cops do their jobs and determine that PINKLADY is not who they're actually looking for, and they go on their way.

Multiple times? Police laziness fueled by AI incompetence

The people getting caught up in this have been pulled over multiple times.


I think if you are driving around in a pink Ford Taurus you are definitely guilty of something even if the plate reads MARYKAY

bad taste isn't a crime lmao

Tell that to the fashion police.

Pigeonholing responsibility onto one party is what allows these mutually-dependent systems to point fingers at one another to escape blame. Rather, the responsibility here is shared. If you want to focus your call for reform on the police (for both making an overly-broad list, and also for harming innocent motorists without compensating them for the damage), then I agree that's more appropriate for this particular problem. But don't absolve Flock.

>Pigeonholing responsibility onto one party is what allows these mutually-dependent systems to point fingers at one another to escape blame

Exactly. The responsibility can't all be pinned on one party and divided no party has enough of it.

Collective guilt needs to make a comeback. Make people and systems have an incentive to associate with malicious or shoddy people or systems.


"They can't remove it without knowing who the warrant is for" is absolutely Flocks problem.

They're alerting on a license plate but yet somehow they can't turn off that license plate alert using just the license plate number? Fucking bullshit


Wouldn't it be the purview of the cops to update Flock that the plate is no longer of interest and to stop alerting on it? I'm no fan of Flock, but let's put the onus where it is deserved.

I can tell you've never worked on government software.

> Flock is just doing what it is being asked to do!

Well then clearly they are not a problem.


Hmm, I wonder what Flock proponents would say when immediately asked about guns, after all, it's just a machine doing what it is being asked to do!

This is precisely what they mean when they say "guns don't kill people, bad people with guns kill people"

What that statement misses is that guns make people bad.

Are we up in arms about gun manufacturers selling guns to the police because police sometimes misuse them?

they'd only support fully AI-driven guns with zero oversight

If Flock flags license plates at the request of the government, it is acting as an agent of the state and is required to meet governement/constitutional requirements.

How does flock get around this? It can't be an agent of the state AND be private and exempts from 4th amendment/all constitutional requirements?

https://www.fletc.gov/audio/definition-government-agent-unde...

Solari: No sir, unless he was for some reason acting on behalf of the government or had been asked by a government agent to do that. Unless that were the case then if that person was acting in his own private capacity as a UPS or FedEx employee then he would not be a government agent for 4th Amendment purposes.

Miller: Can private parties ever trigger the 4th Amendment?

Solari: Yes, as we discussed, if a private party were to be acting at the behest of the government -- if a government agent were to ask that FedEx person to open up a package and look inside, or to ask someone’s girlfriend to go through their things looking for evidence to turn over to the police, then that would be government activity. That would be the actions of a government agent because government agents can’t ask private parties to do something they themselves couldn’t do under the 4th Amendment, so in that type of instance it would be extended to that private party.


Back in the day, CACM use to have job listings at the back. There was always DE Shaw advertising. Sending in printed CVs and cover letters was the slow and painful way to do it.

It will be interesting to see how Goodell's citations drop going forward.

Why doesn't the comet "streak" also, given the Earth's rotation? 10 minutes is a long enough window to have an appreciable impact on the comet's image. Or is it the case that the telescope is stabilized to the Earth's rotation?

Low earth orbit satellites orbit about once per 90 minutes, so in 10 minutes they go about 40 degrees across the sky. The comet is not even orbiting the earth, it's essentially fixed in the sky. The earth only rotates about 2.5 degrees in 10 minutes. So the satellites streak is 16 times longer than the comets.

> essentially fixed in the sky

Let's start with "fixed in the sky" and qualify your frame of reference as the field of distant stars, or the celestial sphere. The common coordinate system is right ascension (RA) and declination (dec).

The GP question was about the Earth's rotation, which would be in terms of azimuth and altitude, and that question's been asked and answered. The key terms there: "equatorial mount" and "clock drive".

The comet C/2025 R3 (PanSTARRS) is in fact near its highest velocity (with reference to the Sun especially), being near perihelion while this photograph was taken. The comet is swinging around the Sun, and it was about 0.49 AU from Earth at the time of the photograph.

https://en.wikipedia.org/wiki/C/2025_R3_(PanSTARRS)

I chose an approximate time on April 27: in 10 minutes of wall-clock time, with the J2000 epoch, the comet's apparent motion is from RA 02h 49m 07.1s, dec +06° 02' 56.5" to RA 02h 49m 15.4s, dec +06° 02' 13.3"

That is a distance of 2' 11.13" across the celestial sphere. For reference, Venus is 11.6" wide in the sky as we see it this week.

24 hours later, we find it at RA 3h 08m 44.1s, dec +04° 19' 27.8". Its apparent motion was 5° 10' 46.02", which is approximately the width of your three middle fingers held together, at arm's length.

So, "fixed in the sky" is not a scientifically useful description of astronomical objects: we need to put that in terms of at least one frame of reference, and "apparent motion" which is how an observer perceives it.

https://soho.nascom.nasa.gov/data/LATEST/latest-lascoC3.html (grab this today; scroll between 4/23 and 4/27)


Given that the moon is about 0.5 degrees in diameter from Earth, shouldn't we expect to see the stars and comet much more blurred than they are here though? Or the ground if it's stabilized against the rotation?

40 degrees around Earth (central angle)

but it increases to much more when you are much closer to the arc


Great point!

https://www.facebook.com/groups/Nachtfotografie/posts/264063... Here is the original photo description in German. See also my other comment in this thread. But the tl;dr is that this was a stack of 153 four-second exposures with some gaps in the timelime when the camera took its time to save between exposures.

On a related note, another question: who owns the paper that Claude (or OpenAI) wrote? Should such paper submissions in conferences call out the model(s) used to write the paper itself?

This is what I've heard on the "street". Building a CUDA-compatible stack for AMD's hardware requires highly-paid SWEs. It's a very niche field, and talent is hard to come by.

But AMD does not want to pay these specialized SWEs the market rate. Their existing SWEs would be up in arms saying, basically, "what are we, chopped liver??", or so the thinking goes.

So AMD is stuck with a shitty software stack which cannot compete with CUDA.

If I were making such decisions, I would just cull the number of existing SWEs down by 50%, and double the pay for remaining ones. And then go out and hire some top talent to build a good software stack.


> highly-laid SWEs

Freudian slip?


Ha! You caught it before I did; and I caught it right away.

Is there anything wrong with using AI (Claude Code/Codex/Gemini etc.) to design your website or your app? As an engineer, I know what my strengths are; and I am pretty damn sure "reactive website design" is not one of them. Why not use AI to do the heavy lifting?

Do you even really need "reactive web design"? Not every website needs to be a webapp, if the sites just a blog and links to the docs and downloads it's very reasonable to sit down and crank out some HTML and CSS like the good old days. Hell, use a static site generator and the HTML bit mostly disappears

Bad software kills good hardware.

And the converse is true also. I mean, look at NVIDIA. For the longest time they were just a gaming card company, competing with AMD. I remember alternating between the two companies for my custom builds in the 90s and it basically came down to rendering speed and frame rate.

But Jensen bet on the "compute engine" horse and pushed CUDA out, which became the defacto standard for doing fast, parallel arithmetic on a GPU. He was able to ride the BitCoin wave and then the big one, DNNs. AMD still hasn't caught on yet (despite 15 years having gone by).


I make the mistake of thinking its 2020 as well. CUDA was announced 2006 and released Feb 2007. So its actually 20 years that AMD/RADEON hasn't caught on that they need a good software stack.

FTA:

> One pod of TPU 8t is 121 ExaFlops; or 121,000 PetaFlops.

Meanwhile, the compute capacity of the top 10 supercomputers in the entire world is 11,487 Petaflops.[1]

I know, I know, not the same flops, yada yada, but still. Just 1 pod alone is quite a beast.

Edit: [1] https://top500.org/lists/top500/2025/11/


Not only not the same size, 4-bit flops versus 64-bit flops, but not the same programmability either. the TPUs can do just matrix-multiplications and some supporting math.

Otherwise bitcoin mining rigs dwarf everything, if you just want to count raw operations per second.


Other than TPUs they're also planning for 960,000 Rubin GPUs [1] which can do 33 teraflops fp64 each, so over 30 classical exaflops, and with emulation it could be more than 100 exaflops.

[1] https://blogs.nvidia.com/blog/google-cloud-agentic-physical-...


0 to $60B in less than 4 years ... impressive!

Consider applying for YC's Summer 2026 batch! Applications are open till May 4

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: