A piece of data on its own is usually pretty harmless, a gigantic database that is organized, categorized, and distributed, is potentially very harmful. And it's a smooth gradient between them.

An example of data that is harmless on its own but harmful at scale is license plate reading. https://arstechnica.com/tech-policy/2015/03/we-know-where-yo... a few points tells you that a person is in a city, a few more tells you where they live, a few more tells you where they work, a few more tells you where the shop, a few more tells you who their friends are, who their associates are, if they're seeing someone on the side, if their going to any gatherings, if they tend to frequent red light districts... ect.

In this way, as data is aggregated, the picture of a given person slides from blurry to clear. And a clear picture of someone is an invasive and dangerous picture of someone.

As a result, aggregation of "public" data like this is something privacy advocates generally resist. It's not that there's is some line that's been crossed.

Exactly. It's the industrial scale collection, storage, analysis, and sharing of data that becomes a problem. I don't care of Joe Smith the security guard sees me pick my nose or scratch my ass when they're watching security cameras because it's not noteworthy to them and they're not going to recognize me or remember it.

However, I do care if the cameras are feeding into a huge system for analytics and data sharing. What if I cough 2x more than the average person? Does that get tracked, stored, and sold to a health insurance data supplier that ranks me as high risk for respiratory problems?

Does the machine learning algorithm that analyzes footage rank me as a potential thief if the database says I hang out in a lot of poor neighborhoods?

These will turn into class profiling systems that will amplify everything in your life so much that whatever your economic status is when you're born is what it'll be when you die.

The answer to all of these questions is: if there is money to be made on siphoning off some data here or there, someone will invent a way to do it. Sometimes silent, sometimes brazen. Usually silent though.

Do you use a credit card? Why are you so upset about this dataset but not that dataset?

Honest question.

I am definitely not happy about that either.

Having a card is just as unavoidable as needing to go to the shopping centre to buy groceries, so this isn't even a case where "if you don't like it don't use it" applies

Yes. And why do you think I'm not upset about that dataset?

You haven’t actually articulated any harm here - the “dangerous” actions that could potentially happen with the aid of aggregated data, like blackmail or stalking, are already illegal.

It feels a little onerous for me to lay out the a privacy advocates argument from first principles.

I'd presume if umvi had rejected the groundwork reasons for privacy, they would have made a comment more like:

"What bad thing are we asserting would happen if they created this data set? What demonstrable harm can you show from even an omnibus dataset? Let alone what this mall is creating?"

The end equilibrium is everyone running cameras constantly in public spaces, and all public activities scrutinized by... well, who knows? For what ever purposes they choose.

As just one example, I'd expect the emergence of semi-automated blackmail - you can infer a lot from peoples' public activity.

"Merchants should be able to hire humans that sit around looking at camera screens all day trying to find shopping patterns, shoplifters, etc."

There is no identification part in what you describe. Generic stats (how many person in the queue, etc.) are okay.

The issue is not facial detection but facial recognition.

They stored model of faces associated with data which mean they could point a camera at anyone in the street anywhere in the world and find back what is the data associated with that person.

Hiring humans would not result in creating and storing models of faces

You're telling me humans can't recognize faces? Consider a security guard who sits at the entrance to a store and watches people come in. What are the chances he would facially recognize a repeat shoplifting offender?

In scenario 1, getting a warrant doesn't do a whole lot of good.

In scenario 2, if the technology is widespread, police can be searching all stores records country wide within minutes.

Yes, because humans looking at camera screens will a) cost the company money, and also b) creates an incentive to the company to limit its observation to only the most necessary viewpoints.

Attempting to automate the task of constant observation makes constant observation easier, which isn't necessarily a desirable state.