• 0 Posts
  • 1.05K Comments
Joined 8か月前
cake
Cake day: 2024年12月6日

help-circle




  • You’re still thinking it at a the level of “can”, rather than the level of “is it worth it”.

    “Possible” isn’t the same as “profitable” - the whole point of stores doing it would be to sell that data to entities such as Health Insurers, and that data would be sold at pennies (and I don’t mean pennie-per-entry, I mean pennies-per-thousands), so it has to be possible to do it extremely cheaply, which it is if you already have the necessary information in digital form in a database (user id from the payment card and the list of items purchased along with date, time and location of purchase), but it’s not if you have to do reverse image search in bulk, not least because the providers out there won’t just allow other businesses to do it for free to make money out of it - they’ll demand a cut for doing the computationally hardest part of the process (or, if the supermarkets want to do it themselves, for access to their store of pictures).

    Also, the quality of results from reverse image search is pretty bad in terms of actually finding and correctly identifying a person from a picture - it often just outright fails or gives false positives, which means data obtained that way is a lot more polluted with false results than just finding the person ID via the card used for payment, which is near perfect (not quite perfect because somebody might let somebody else use their card or the card might have been stolen, but way more reliable than identifying somebody via a picture).

    So all this hassle and cost to have a parallel process to try to ID people like me who pays in cash to sell my purchasing habits information, when most people are like you and just give them their ID on a platter by paying with card, doesn’t make any business sense.

    They ain’t doing it because they can’t, they ain’t doing it because so long as the market is flooded with customer purchasing habits data obtained cheaply by just using the information from their card payment, deploying facial recognition technology to match buyers to purchases wouldn’t actually be profitable.

    Just because it’s technologically possible to go after the hard to get info using a complex process, doesn’t mean it makes business sense to do it, especially when they’re already making money with a far simpler and cheaper process.


  • I suspect that outside a well controlled environment (like that small tunnel with a lot of cameras), image recognition still yields too many false positives and false negatives to be acceptable compared to scanning a bar code (then again, maybe scanning barcodes is what that tunnel does rather than image recognition).

    That said, there was this whole idea of using RFID tags on products so that checking-out was merely passing by a scanner with your filled trolley - which would scan all of its contents at once - and paying (or even have your card directly charged).

    However I believe this failed to take off because neither product manufacturers nor the stores wanted to spend the few cents per box that would take to add the RFID tags.

    So in order to save the few cents per-box that would enable pretty much instant checkout, we have these crap self-checkout implementations were clients get to do all the work of cashiers in a teller which is worse than that of cashiers, and without even getting a discount for it (actually prices just kept going up) - the whole thing is fucking insulting.



  • Neural Networks, which are the base technology of what nowadays gets called AI, are just great automated pattern detection systems, which in the last couple of years with the invention of things like adversarial training can also be made to output content that match those patterns.

    The simpler stuff that just does pattern recognition without the fancy outputting stuff that matches the pattern was already, way back 3 decades ago, recognized at being able to process large datasets and spot patterns which humans hadn’t been able to spot: for example there was this NN trained to find tumors in photos which seemed to work perfectly in testing but didn’t work at all in practice, and it turned out that the NN had been trained with pictures were all those with tumors had a ruler next to it showing its size and those without tumors did not, so the pattern derived in training by the NN for “tumor present” was actually the presence of the ruler.

    Anyways, it’s mainly this simpler and older stuff that can be used to help with scientific discovery by spotting in large datasets patterns which we humans have not, mainly because they can much faster and more easily trawl through an entire haystack to find the needles than we humans can, but like in the “tumor detection NN” example above, sometimes the patterns aren’t in the data but in the way the data was obtained.

    The fancy stuff that actually outputs content that matches patterns detected in the data, such as LLMs and image generation, and which is fueling the current AI bubble, is totally irrelevant for this kind of use.


  • Clearly you never actually done Tech projects in large corporate environments if you think complex shit is implemented across all sites just because it can be done, rather than because the expected profits exceed the cost and the hassle.

    Also you seem to be under the impression that the social media guys would just give searchable access to their store of pictures (or provide a search service) to those big companies for free, which is a hilariously naive take on how Tech businesses work.

    Automated following customers in a store with overhead cameras for the purposes of studying how they move around and purchase things is only done for some stores and has entirely different requirements for camera positions, external dependencies (no cross-referencing with external data to ID anybody is needed) and acceptable error rates (the data is not for selling to others so the error rates can be higher), because they don’t need to actually ID anybody to extract “human movement patterns” out of that data and it’s fine if the system confuses two people once in a while because there is no external customer of that data getting pissed off when the same person is reported as making purchases in two places at the same time or other stupidly obvious false positives.

    Meanwhile matching the list of items bought with payment information, both of which already get sent from the tellers to the backend systems (for purposes of inventory tracking and accounting), is easy peasy and has a very low error rate.

    You’re ridding a massive Dunning-Krugger there in thinking you’re the expert.


  • There’s a good argument to be made from the point of view of customer convenience and expediency for using self-checkouts to pay for a small number of items, but even then most existing implementations of that concept are so fucking bad that there are all sorts of stupid problems, like my case of the thing not working unless I had a bag (it literally had no button to just skip it) and yours were a normal human mistake is complex to correct even though the users are amateurs and hence naturally more likely to make mistakes hence the thing should have been designed differently.

    I’ve actually worked with UX/UI designer at several points in my career, and one thing that pisses me off about most self-checkouts is just how bad their UX/UI design is.

    That so many self-checkout implementations are like that is probably explained by, having moved the costs of wasting time to the side of client, those businesses are not financially incentivized to make the self-checkouts efficient to use, which probably also explains all manner of weird choices in everything from their shape to even the order of their menus - in a manned checkout it’s their problem because wasted is money being paid to a teller for nothing, so if it’s bad they fix it, whilst in self-checkouts it’s not their problem so they don’t care.

    This is also another reason for me to be against self-checkouts: the financial dynamics are different with self-checkouts than with manned checkouts since the costs of inefficiency on the former are on the customer, whilst with the latter the costs are on the store (which has to pay a salary for somebody who is less productive than they could be), so stores have less (and more indirect, hence harder to measure, hence often ignored by MBAs) financial pressure to make self-checkouts efficient to use than they do with manned checkouts.


  • They have to go massively out of their way, spending a lot more more money both in hardware and ongoing processing power costs, to do that kind of tracking which gives far less reliable results, than simply matching the entry in the database of a specific purchase with the person identified by the card that paid that purchase.

    Your “argument” is akin to a claim that people shouldn’t worry about having a good lock on their door because it’s always possible to break the door down with explosives.

    “Don’t be the low hanging fruit” is a pretty good rule in protecting your things, including protecting your privacy.

    But, hey, keep up the good work of giving them all your personal info on a platter so that their ROI of investing in the kind of complex tech needed to do tracking of people like me remains too low to be worth it.


  • Mate, not the previous poster but I’m a senior software engineer with an EE degree and broad enough experience that I could design and implement myself a self-checkout from the ground up, both hardware and software, including UI and backend integration, and I still tend to avoid self-checkouts for those reasons and a lot more (many which I listed in another post here).

    There are two very opposite ends of the curve for people who don’t like self-checkouts: those who can’t deal with the tech (who you deem “fucking morons”) and those who have evaluated self-checkouts as a process and found it to overall be inferior to the existing process for their own usual use conditions or who look at it in a broader context and find it to have indirect social damage.

    That you can only spot the “being a moron” as a reason to avoid self-checkouts is a pretty good indicator of your own intellectual limitations.


  • Further:

    • Most self-checkouts are too small and unwieldy to hold two shoppings bags when you’re packaging a week worth of purchases.
    • You still need an employee to come over and certify that you’re over 18 if you buy alcoholic drinks, and there’s usually just one for many tills who is usually busy with somebody else.
    • I like to pack my weekly shopping in specific ways (cold items together, fragile stuff on top, weight balanced) and whilst in a normal checkout I can do packaging in parallel with somebody else doing the checkout plus already place things roughly ordered on the threading band to the cashier, in the self-checkout it’s just me and things are in whatever order it went into the trolley so it takes at least twice as long.
    • They often have quirks, such as for example the one I used more recently would not let me start unless I put a bag in the output compartment first, so I needed to have or buy a bag even though I was buying just 1 item (mind you this might have just been trying to force people to buy a bag, since many forget to bring one - in other words, structuring the software to force people to spend money which is a form of enshittification).
    • They’re non standard and each store has a different model, with different physical structure and different software with a different UI with buttons in different places and often different quirks, so anything you learn beyond the basics about how to use one effectively is often non-translatable to self-checkouts in different stores.
    • They often don’t take cash. Cash is good, it means your buying habits are not in some database somewhere and used for things like having an AI estimate how much an airline company can wring out of you for a ticket for a flight or a Health Insurer assessing your risk profile and upping your price, it works always even during outages (of power, of your bank, of payment processors) and studies have shown people save money if they pay in cash because they tend to spend less (something about the physicality of parting ways with your notes and coins makes people be more wary of paying more than if it’s just a number on a screen).




  • It depends on the country.

    Germany “unwaveringly supports” it (it’s tradition for them to support large scale extreme Racist Genocides, I guess) as does Italy (though there it seems more of a “brotherhood of Fascists” thing between their ruling coalition and the Zionists), whilst Spain and the Republic Of Ireland condone condemn it, and a bunch of others like The Netherlands and France make some noises of discomfort about it but still keep on selling weapons to Israel.

    There really is no EU-wide agreement on this, though most of its member countries are slowly shifting against Israel.