Live data from Hacker News

Clearview AI: secretive company that might end privacy as we know it (2020)

nytimes.com

61–70 of 76 posts

Re: Clearview AI: secretive company that might end privacy as we know it (2020)

#61
post #33

Earlier quoted context omitted.

Scary thought: Let's say you do this. And facial data DBs continue to grow. And eventually they become so widespread that most people are in them. And facial recognition becomes so ubiquitous that it is necessary for everyday transactions. And you can't participate in large swatches of society because you are, as far as the technology is concerned, a statue, or dead. Far fetched? It's happening with cell phones. Many…

Good! A 0.01 - 1% failure rate and people will become very pissed off at the system, making it politically untenable. Which for a small country means cloning and poisoning the internet for 200 - 20.000 people. Not too impossible.

>A 0.01 - 1% failure rate and people will become very pissed off at the system, making it politically untenable

I'm quite sure I don't live in a society where something (even greatly) inconveniencing a mere 0.1% of the public makes it politically untenable, and I'm very nearly as sure that you don't either.

Re: Clearview AI: secretive company that might end privacy as we know it (2020)

#62
The right to control our public exposure is just one facet of the right to privacy.

The right to privacy is personal sovereignty. Meaning control over oneself.

I am my data, my data is me. I own all data about me.

Recognizing our collective right to privacy prohibits facial recognition used for mass surveillance.

Firstly, each person must consent to being represented in those datasets.

Secondly, each person must consent to being tracked.

Have you given consent? I certainly have not. I wasn't even asked.

Re: Clearview AI: secretive company that might end privacy as we know it (2020)

#63

Clearly the countermove is to train an image generator on your face and flood the internet with fake data. Have your face appear tagged in weird remote places, with several made up names, on statues, in ancient manuscripts. Pretend to be dead, pretend to be a medieval monk, a war hero from the 1940s, a fictional character. Poison those ratty algorithm with nonsense and hide within the noise.

The countermove is what D-ID (YC funded, IIRC) said they would do and then changed their mind: train a face recognition library to generate modified versions of a picture such that a human would still believe it's the same person, but different enough that face recognition systems think it's someone else. Then use those pictures in all of your internet socials. Edit: according to this article they did develop that, b…

Facial recognition algorithms are a set of different and developing models. There's no guarantee any system aimed at the weaknesses of an existing algorithm would work against different or future facial recognition systems.

Re: Clearview AI: secretive company that might end privacy as we know it (2020)

#64
post #5

Pretty sure that pretty much all of their business practices are GDPR violations. So we do have a choice - we could either end privacy, or have legislation that ends a potentially profitable startup.

"Won't someone think of the employees" is a common tactic companies use to avoid shutdowns, and that's against long standing laws and regulations. Seems to work often enough too.

[dead]

Re: Clearview AI: secretive company that might end privacy as we know it (2020)

#65
post #25

>One reason that Clearview is catching on is that its service is unique. That’s because Facebook and other social media sites prohibit people from scraping users’ images — Clearview is violating the sites’ terms of service. “A lot of people are doing it,” Mr. Ton-That shrugged. “Facebook knows.” So I own the copyright to the image I uploaded to facebook I licensed it to Facebook they stole it from Facebook. Did they…

In Europe: no. Training ML systems doesn't require a license. In the US: we don't know yet. The law is not clear, and there are lots of suits in progress.

You misunderstand. They copied the literal image from facebooks computers to their own to train with.

Re: Clearview AI: secretive company that might end privacy as we know it (2020)

#66
post #25

Earlier quoted context omitted.

In Europe: no. Training ML systems doesn't require a license. In the US: we don't know yet. The law is not clear, and there are lots of suits in progress.

You misunderstand. They copied the literal image from facebooks computers to their own to train with.

And that may be permitted under fair use. It is not yet settled law in the US.

Re: Clearview AI: secretive company that might end privacy as we know it (2020)

#67

Clearly the countermove is to train an image generator on your face and flood the internet with fake data. Have your face appear tagged in weird remote places, with several made up names, on statues, in ancient manuscripts. Pretend to be dead, pretend to be a medieval monk, a war hero from the 1940s, a fictional character. Poison those ratty algorithm with nonsense and hide within the noise.

The countermove is what D-ID (YC funded, IIRC) said they would do and then changed their mind: train a face recognition library to generate modified versions of a picture such that a human would still believe it's the same person, but different enough that face recognition systems think it's someone else. Then use those pictures in all of your internet socials. Edit: according to this article they did develop that, b…

I want someone to make the ugly shirt.

Re: Clearview AI: secretive company that might end privacy as we know it (2020)

#68
post #21

Earlier quoted context omitted.

The answer is pretty simple: Privacy matters more. There are millions of ways for profit but you only have one privacy.

Privacy doesn't matter and GDPR is not a solution.

Do you live in a glass house?
Post reply on HN