This should be very familiar to people working with data in a lot of jurisdictions. I can speak to Europe but I think similar things exist elsewhere - data is less restricted in how and what you collect than it is how you use it. This makes a lot of sense, you should be able to have a basic record of ip addresses and access times for rate limiting, but that shouldn’t mean you can use it for advertising.
Similarly it seems reasonable that shops should be able to record for some purposes but not all.
I don't think "less restricted" is a good framing. How you are using it is the core, and you get to collect and store what's necessary for your legal uses, and use it for those uses.
You don't get to have access logs because there is no restriction on logging IPs, you get them because you argue a justified use of them, and thus you can have them to use them for it (and not for anything else).
I know what you mean but read this in context. You're less restricted in what you can collect compared to what you can do with it - any valid use case requiring video footage allows you to get video footage but that doesn't mean you can then do anything you want with it. The key is what are you using the data for.
I don't think it does, because it is completely unverifiable. It's like allowing people to buy drugs, but not to use them.
I'm not worried about people collecting IPs, I'm worried about people who collect IPs being able to send those IPs out and get them associated with names, and send those names out and be supplied with dossiers.
When they start putting collecting IPs in the same bag as the rest of this, it's because they're just trying to legitimize this entire process. Collecting dossiers becomes traffic shaping, and of course people should be allowed to traffic shape - you could be getting DDOSed by terrorists!
edit: I'm not sure this comment was quite clear - it's 1) the selling of private, incidentally collected information by service providers, and 2) the accumulation, buying, and selling of dossiers on normal people whom one has no business relationship that is the problem. IPs are just temporary identifiers, unless you can resolve them through what are essentially civilian intelligence organizations.
Don't the industry-imposed rules for handling credit cards work that way (restricting use of data you already have) though?
Like, I thought a big part of why some stores do loyalty cards is because they enable tracking things that they'd get their credit card privileges revoked if they tracked that way.
> 1) the selling of private, incidentally collected information by service providers, and 2) the accumulation, buying, and selling of dossiers on normal people whom one has no business relationship that is the problem.
But this is exactly what is covered - incidentally collected information cannot be used for other purposes. That's rather the point - you must collect things for a specific use case and you can't use it without permission for other cases.
> I don't think it does, because it is completely unverifiable.
It's no less verifiable than "don't collect the data", and hiding it requires increasingly larger conspiracies the larger organisation you are looking at. People are capable of committing crimes though, sure.
You forget store. This depends a lot on the type of data. Duration, specific laws related to it and protection are very different for randomised numbers vs medical as an example.
This is good. It means we have laws and rulings that understand the technology. That balance the need for business to protect their stores with people's privacy.
Free to record data but not free to process data... sounds a lot like books being stored rightfully but not analyzed by machine learning.
I have data on Google. Google has a TOS that says they can use my data. This could cover even future use cases, even though those future use cases I did not anticipate. So does Google have the right to use my data in this particular way?
There are all manner of things you can and cannot do with 'data'. For example, you cannot purchase a Blu-Ray, rip its contents and post them on the internet. This shouldn't be that "interesting".
I noticed that as well, it's a bit frustrating. I personally think if you're allowed to do something legally, you should be allowed to do it using technology.
It's seems silly to me that you can have a human being eyeball someone and claim it's so and so, but you can't use incredibly accurate technology to streamline that process.
I personally don't like the decay of polite society. I don't like asking a worker for a key to buy some deodorant. Rather than treat everyone like a criminal, why don't we just treat criminals like criminals. It's a tiny percentage of people that abuse polite society and we pretend like it's a huge problem that can only be attacked by erecting huge inconveniences for everyone. No, just punish criminals and build systems to target criminals rather than everyone. If you look at arrests, you'll see that among persons admitted to state prison 77% had five or more prior arrests. When do you say enough is enough and we can back off this surveillance state because we're too afraid to just lock up people that don't want to live in society.
Target the criminals. Yes, the criminals, according to the altruistic government that serves its people and never breaks the constitution or international law. According to its perfectly defined code. Someone who got an abortion in another state after rape, for example. Or said "children shouldn't be murdered"? but said it out loud at a campus. Run the tech of the most powerful trillionaires on the masses, the poor people and find out who is dissenting. Keep the prisons full. One man's prison is another man's pension. Make this system more powerful.
We are all potential criminals under tomorrow's government. Remember that!
If you flag the dozen or so people that come in to your business once a week to steal, you don't have to have as much surveillance in the store otherwise. Just check them out when they enter, very simple.
For instance, Costco has a much lower theft rate (0.11–0.2% of sales) compared to other supermarkets (1-4%) simply because they manage to keep criminal out through membership fees. Control the entrance, target the known criminals and we can go back to a high trust society.
First paragraphs pretty clearly read to me like the issue isn’t “processing it,” it’s the indiscriminate filming of everybody who enters the store without consent that’s the problem.
Security filming is common in Australia and not outlawed by this ruling. It is specifically the non-discriminate use of facial recognition technology this ruling applies to.
The specific difference is "sensitive information". General filming with manual review isn't considered to be collecting privacy sensitive information. Automatic facial recognition is.
The blog post makes this point about how the law is applied:
> Is this a technology of convenience - is it being used only because it’s cheaper, or as an alternative to employing staff to do a particular role, and are there other less privacy-intrusive means that could be reasonably used?
I don't really understand their reasoning behind the "technology of convenience" point.
Say I implement facial recognition anti-fraud via an army of super-recognizers sitting in an office, watching the camera feeds all day (collecting the sensitive information into their brains rather than into a computer system). It'd be more expensive and involve employing staff (both the "technology of convenience" criteria. From a consumer perspective the privacy impact is very similar, but somehow the privacy commissioner would interpret this differently?
Maybe that is the point the privacy commissioner is trying to make, that collecting this information through an automated computer system is fundamentally different than collecting this information through an analog/human system. But I'm not sure the line is really so clear...
In the KMart case, it would not have been interpreted differently if people were doing the facial recognition rather than a computer. The issue was indiscriminate use on everyone who walked in, without permission or proper notification. Which is only cost effective if automated, and a technology of convenience I guess.
But is a non-indiscriminate, privacy friendly solution possible? The problem is people walking in with a valid receipt for a purchased item, grabbing a matching item off the shelf, and wandering over to the returns counter and requesting their money back. The usual solution most shops use is locating the returns counter past the security controls (checkout counter). But more and more of these types of stores are putting their service counters in the middle of the store for some reason.
It's a false equivalence to equate humans (even "super-recognizers") with a computer when it comes to matching large quantities of faces with names/PII.
At some point the numbers get big enough that you wouldn't be able to get the pictures of faces in front of the people who would recognize them fast enough.
Everyone who enters almost any store is "filmed" with their implicit consent. Cameras are everywhere, and certainly are everywhere in every Australian court as well.
The root comment is precisely right. Deriving data from filmed content -- the illusory private biometric data that we are leaving everywhere, constantly -- is what the purported transgression was.
Is this from the 90s? Who doesn't expect to be recorded when entering a retail chain? How the hell does the government have the right to decide what this private company can do on their private land? If you enter onto someone else's property you should play by their rules.
People here might be interested in Zennioptical's ID Guard technology, if they wear glasses. Evidently it's not perfect, but it does at least partially work: https://youtu.be/HOBdJ6nU03o?si=E_a6rMPAz5AOwytm
Business opportunity: sell covid masks with patterns designed to thwart facial recognition on them.
Why are they covid masks anyway? Medical personnel wears them during surgery, and there were those photos of ... some asian people i think ... wearing them outdoors to protect themselves from air pollution in their city too.
> How the hell does the government have the right to decide what this private company can do on their private land?
Unless you think a grocery store should be allowed to grab you and sell your organs then you agree that this private organisation should be subject to some limitations about what it can do on its own land. The question is then where the line should be between its interests and the interests of those who go on the land.
You can be absolutist about this, that’s certainly a position, but it’s extremely far from mainstream.
Exactly. There is a limit to what a private company can do on private land, set by "the government" (here it'd be parliament). You don't seem to be an absolutist about this, so we both agree that the government can and should tell private businesses what they can do on private land. Then the issue is only where the line should be not whether there should be a line at all.
> How the hell does the government have the right to decide what this private company can do on their private land?
Because the world is bigger than just the wishes of private businesses. I don't think there is anywhere on this planet where you as a private business can do literally whatever you want, there are always regulations about what you can and cannot do. The first thing is usually "zoning" as one example, so regardless if you own the land, if it isn't zoned for industrial/commercial usage, then you cannot use it for industrial/commercial usage.
What libertarian utopia do you live in that would allow land owners to do whatever they want?
The authors largest bullet point is on author identification. What are the arguments against this and do people really feel they outweigh the benefits?
My view: I would think given how many code supply chain attacks we've see recently this would be be regarded as (at worst) a necessary evil. How much software used by large numbers of people does the open source community think will be done by anons?
Sidenote: The author implies SyncThing development was stopped due to author ID but the post linked does not say this and gives a completely different reason (forced updates)
Papers make specific conclusions based on specific data. The paper I linked specifically rebuts the conclusions of the paper. Gary makes vague statements that could be interpreted as being related.
It is scientific malpractice to write a post supposedly rebutting responses to a paper and not directly address the most salient one.
This sort of omission would not be considered scientific malpractice even in a journal article, let alone a blog post. A rebuttal of a position that fails to address the strongest arguments for it is a bad rebuttal, but it’s not scientific malpractice to write a bad paper — let alone a bad blog post.
I don’t think I agree with you that GM isn’t addressing the points in the paper you link. But in any case, you’re not doing your argument any favors by throwing in wild accusations of malpractice.
Many people drive places other (further) than work multiple times a year. "75 mile battery" wouldn't even be good enough for a one-way trip of this kind let alone there and back again.
People are drawing erroneous conclusions from this.
My read of this is that the paper demonstrates that given a particular model (and the problems examined with it) that giving more thought tokens does not help on problems above a certain complexity. It does not say anything about the capabilities of future, larger, models to handle more complex tasks. (NB: humans trend similarly)
My concern is that people are extrapolating from this to conclusions about LLM's generally, and this is not warranted
The only part about this i find even surprising is he abstract's conclusion (1): that 'thinking' can lead to worse outcomes for certain simple problem. (again though, maybe you can say humans are the same here. You can overthink things)
You can absolutely extrapolate the results, because what this shows is that even when "reasoning" these models are still fundamentally repeating in-sample patterns, and that they collapse when faced with novel reasoning tasks above a small complexity threshold.
That is not a model-specific claim, it's a claim on the nature of LLMs.
For your argument to be true would need to mean that there is a qualitative difference, in which some models possess "true reasoning" capability and some don't, and this test only happened to look at the latter.
The authors don't say anything like this that I can see. Their conclusion specifically identifies these as weaknesses of current frontier models.
Furthermore we have clearly seen increases in reasoning from previous frontier models to current frontier models.
If the authors could /did show that both previous-generation and current-generation frontier models hit a wall at similar complexity that would be something, AFAIK they do not.
I guess the authors are making an important point (that challenges the current belief & trend in AI): adding reasoning or thinking to a model (regardless of the architecture or generation)doesn’t always lead to a net gain. In fact, once you factor in compute costs and answer quality across problems of varying complexity, the overall benefit can sometimes turn out to be negative.
Human brains work the same way. Some of us are just better at analogy. I’ve worked with plenty of people who were unable to transfer knowledge of one area to another, identical area with different terms.
That isn't the reason. They sunset in the bill so it has a lower CBO score (which calculates costs out to 10 years). If you sunset in the bill after 5 years, even if you know it will get renewed, the apparent cost goes down. Get it?
Profanity should not be in the title of scientific articles. Most unprofessional. In addition, titling your article for shock value should be discouraged. The end point will be a degraded discourse.
I've joined jobs and the first thing people said to me is "ah, you must be the new cunt!"
Different people have different standards for this type of thing. Be a good cunt and accept that there are over 8 billion people on the world, some of whom have very different norms than you have. Don't declare your own standards as somehow authoritative.
No, you have certain ways you like communication to happen. That's okay, everyone had that. To present this as some sort of objective standard is complete bollocks, as is your claim that it somehow "degrades" discourse.
This applies twentyfold when the topic of the scientific paper is swearing. Like mate, seriously?
I tend to agree. A lot of medical and scientific writing often falls on deaf ears because most people only respond to a conversational tone. That's why you write corporate emails in a conversational tone, it's just what's most effective.
I think, if the subject matters call for it, which clearly this does as they're literally looking at swearing, then it can be fine to swear. It can be more concise and more accurate.
I agree as well, I really dislike the overly formal tone we've tended to adopt in order to signal that the content is important. If you have important stuff to say, it'll be important even if you use simple words to say it.
We tell children to don't use profanity because they have a hard time regulating themselves. Telling adults to do the same is misplaced authoritarian behavior, the kind that may come from people who failed to mature and still obey (and repeat) what they were told as a child but now sound obsequious.
Profanity largely exists to be offensive, and loses power when ubiquitous. It especially loses power when five year olds say it for every little mood swing they have.
Nobody wants to hear offensive words from a child because it makes adults realize how childish they sound.
I suspect the use of profanity was to grab people's attention, rather than for shock value. I would consider it as unprofessional, much as I would consider an article titled "Stars that go boom" to be unprofessional. I would suggest that it should be discouraged, mostly because we don't want scientific journals to come off sounding like tabloids. Yet I don't think that it automatically results in degraded discourse.
TSMC's approach here sounds sensible but I don't think it speaks much to QC. It is a pretty different problem domain. The trapped-ion QCs can use much more expensive / less practical lasers and optics and still be useful.
I'm in a similar boat. Does anyone know any US PCB assemblers similar to PCBWay or JLCPCB? (e.g. small volume) I couldn't find any when i looked about a year ago
We work with a few, but they just do assembly, you bring the parts and PCBs.
They are also insanely expensive if you are used to China prices. Boards with 10-15 components being $100+ each. Couple that with a domestic PCB manufacturing which is ~10x the cost of China, you are looking at something like an Arduino costing $250.
I don't know where you are located, but if you just search google maps for "electronics assembly" you can find places to call. You really want something nearby because it is pretty much the only upside that you can go and talk to them while going over the board and you don't have to hassle with shipping. You can also go inspect the first few boards done before they do the whole run.
I mean in this case the upside for me is no tariff. I don't mind paying some premium for assembly in USA.
Perhaps I'm missing something, but I haven't ever had to 'go over' the board with PCBWay. Maybe why they are so cheap, but I've only ever had one quality issue.
I send the gerbers and pick and place and take it from there. I approve photos If they have any question (which is rare) its over email. I don't even mind sourcing the parts/pcb/stencil myself. Its the components which make the tariff painful.
I used OSH PARK ~ 10-12 years ago - it looks like they're still around and somewhat affordable, but that was just a 20 second scan of their homepage. They used to be the affordable choice back in the day.
If I remember correctly they were just over $200 per smartwatch. We saved some costs by putting the assembled boards into 3D printed cases ourselves using student labor:
It was a double sided, 4 layer board with two small sub-PCB's (touch pads and a power board I think). My experience with Emerald Technologies (then called DataEd at their New Hampshire location, seems they've been bought out by a larger firm) was very positive. They were quite helpful in bringing the project to completion.
The cost per board would have been less if we had ordered more of them. Ours was a fairly large job for DataEd at the time, though they would have been happy to make a thousand of the devices. For tens of thousands or more they would have recommended another affiliated company. They were more of a prototyping company, though that may have changed since the merger. I see that part of Emerald is located in China now, though the DataEd facility is still part of the company, so they might be interested in some US work to carry them through the tariff chaos.
Edit: I should mention, I supplied some of the parts that I prebought to lock in a supply, the MSP430's and touch sensor IC's for example, and they supplied the rest. We had them do minimal QA on the boards, without firmware loaded and we did final QA after programming the boards ourselves. They would have been willing to do those tasks also though.
- you can record all manner of video in your store...
- but you can't process it in this particular way.