Hacker News Reader

Comment by Aunche | Hacker News Reader

I genuinely don't understand what the big deal is. Brick and mortar stores lose a lot lose of money in theft. Walmart alone loses $3 billion each each year. It's their private space, so it makes sense that they want to know who's going inside. The majority of shoppers are using their credit card and store rewards card anyways, so they already know who most of the paying customers are anyways.

View on HN

For me the issue is that it was apparently done in secret. The difference between CCTV that's stored for a day or two and then permanently deleted and cameras tracking behavior and recording data forever is different enough to require a very different kind of disclosure / signage.

"We track customer movement, behavior and tie that to an individual customer record. We keep this information in perpetuity to serve you better."

Gotta have that "serve you better" on there.

Not only that, usually cctvs have a tv facing the entrance with the feed and a reminder that you are on camera. This on the other hand, is Orwellian in comparison.

Would you think it's a big deal if the cops pulled you over, and arrested you for grand theft, took you into booking, finger-printed you, took your picture, etc, then tossed you into a jail cell for a few days because it was a weekend and no judge was available, so you couldn't post bail?

Even after you go through the rigamarole of proving your innocence, you still have to live with having been roughed up by the police, having your mug shoot plastered all over the internet, and having experienced county jail first hand.

And that assumes that you're not charged with something else that happened during the course of the arrest.

The problem is that these systems are deployed in communities such that they implicitly target black people, yet plenty of research shows that they aren't even remotely effective at identifying black people, at all. And the criminal justice system doesn't care.

The criminal justice system doesn’t care what race you are. It is an equal-opportunity devourer.

And that is what makes it scariest of all.

The laws on our books might be race-blind, but the original intentions by lawmakers, enforcement by police and prosecutors, and behavior of juries are most definitely not.

“Racism” is one of the most loaded concepts, but it’s a fact that even by this incredibly loaded definition the system is more racist than any other system in all of human history.

But it isn’t racist in the way that you expect it to be. It isn’t racist at the level of the intersection of the system and the “street”.

The problem that you think is the problem, isn’t. The real problem is much more insidious. And if you attempt to solve the problem that you think is the problem you will do nothing but worsen the problem, just as all previous interventions.

It is what it is.

Sure, but it's still not the core issue. If racism were magically solved overnight, very little would change. You'd still have these corrupt systems exploiting the weak and poor.

Just so.

It most certainly does care, Black men get on average a 20% longer sentence for the same crime as white men.

https://www.ussc.gov/research/research-reports/demographic-d...

Given that the legal system targets based on gender more than race, why race is the go to inequality that is pointed out? There seems to be a social level acceptance that unequal targeting based on gender is somehow more acceptable (and by targeting I mean everything from the initial profiling and surveillance all the way to the chance of conviction and the sentence handed out, with each step granting compounding interest on inequality).

If facial recognition were 99.9% accurate, it would confuse someone with at least 350,000 other people in the US.

If it's not useful, companies aren't going to pay money to use it. As Rite Aid put it, "companies seem to be scaling back or rethinking their efforts around facial recognition given increasing uncertainty around the technology’s utility."

The companies don't care about a few false matches here or there. But if you're one of the people it keeps incorrectly identifying as a criminal who needs to be escorted out of the store, it will matter to you.

Companies do all sorts of insane, crazy things that waste inordinate amounts of money with very little return on investment.

I wonder if it's simply not working as well now that everyone is wearing masks.

Sort of but not really because it's 1/1000 times the chance that they shop at stores in the same geographic location. So yeah I might have a couple dopplegangers but they're likely spread around the US and can be filtered out relatively easily.

It only becomes a problem if I (1) have a doppleganger, that (2) likes to steal, and (3) lives in the same area as me. It'd still be a kafkaesque nightmare but probably a pretty rare one.

Additionally there's other identifiers that when coupled with the 1/350k, they can figure it out pretty quickly.

Throw in a face mask and this tech is all useless. Especially now that we’ve normalized mask use in the U.S. for probably any cold and flu going forward.

They can just do gait recognition as well.

That's why you always shop at Walmart with a roll of pennies in a shoe.

Not if you ride the mobility scooter.

You don't have to be walking. I recently met someone that I had not seen for a long time, recognized them by their pattern of small head movements.

Ok just brainstorming here: mobility scooter, neck brace, leg braces, surgical mask and gloves, sombrero (maximum coverage) and aviator sunglasses.

add on the ill fitting clothes and perhaps those funky sunglasses with the eyeball holograms.

A better way to think of it: if the system only has a 0.1% chance of an incorrect match and at least 700 people regularly shop there, each of them is more likely than not to be misidentified as somebody else.

Think of it more like browser fingerprinting & bits of entropy instead of a one-shot classification. E.g. cross-reference with other data to get near-100% accuracy:" digital fingerprints being emitted from your phone, tracking movement, checking out with a credit card, license plates entering the parking lot, home address vs store location, etc.

They'll have have access to all your store visits, what your family and friends look like and their store visits, what they buy. And a manager/tech company/security company who may or may not be poor/struggling and VC funding deprived/politically motivated will have at least all this a few clicks away for your lifetime.

Sure they have the right to this, and you can choose whether or not to shop with them if they do, but would you want them to without your consent?

I'm probably OK with it as long as ALL OF the following apply:

(1) Only used within private property. No cameras pointing out into the streets.

(2) Used for security purposes only, not marketing or tracking the customers and selling a profile of them to other organizations.

(3) Adequate IT practices so that access is restricted to only employees who need it and so it isn't hacked into and used for whatever nefarious purposes.

(4) Deployed based on objective criteria based on real data, like loss (theft) rates of stores. Not some regional manager looking at a map and deciding that these stores get it and these others don't based on their subjective idea of what qualifies as a good neighborhood and a bad neighborhood.

(5) Technology can be shown not to treat one group of people worse than another. You don't want scenarios like a system that, when dealing with people of one race, is bad at differentiating individuals, so that a legit ban on a person of that race results in a higher false positive rate for others of that race.

For #2 and #3, it would probably help to just not record (or have very limited retention) of the data.

For #4 and #5, some of this might already be legally required already in the US because of existing civil rights laws that prohibit denying customers access based on race, etc.

I agree! I also love how the article says it's "racist" which stores they went into first. Rite-Aid said "Which stores have the most theft?"

I believe that makes for implicit racial bias.

E.g. a person could go to a store and either Find or Not Find a camera. For stores where they Find a camera, they can choose to form an opinion about cameras and whatever 'races' shop here, and leave.

E.g. a person could go to a store and see they are on camera. They can make an association about anti-theft cameras if the store is dominantly in a neighborhood of a specific 'race'. Then they go to another neighborhood dominated by a different 'race', from the same Store Chain, and see they are not on camera. They can speculate this neighborhood is trusted more by the store to be lower in theft.

E.g. two people could then be shopping in one store which has no cameras. One person belonging to a 'race' where cameras are typically found is seen by another. The other assumes, based upon prior observation, that this person of that 'race' is associated with theft. They are suspicious of that first person. ==> Bias against a person of a different race.

That's less than 1% of Walmart's total sales. It's not a major loss for them. Especially when Amazon is offering many products at a much cheaper price-point.

I don't support either major retailer, I'm just providing context that within the given sales (and margin), it's barely a drop in the bucket to a problem for Walmart.

1% of sales for Walmart is ~3-4% of profit. That's huge.

Dude, 1% of sales is absolutely massive, especially as Walmart's profit margin (I assume), isn't the fattest.

>> 1% of sales

It's 0.57% of sales [1]. I tried to find out how much in returns they deal with but...I suspect that returns is a higher number or at minimum comparable.

[1] - https://www.supermarketnews.com/retail-financial/walmart-tur...

Reply on news.ycombinator.com