Slashdot Mirror
IBM, and Some Other Companies Did Not Inform People When Using Their Photos From Flickr To Train Facial Recognition Systems (nbcnews.com)
IBM and some other firms are using at least a million of images they have gleaned from Flickr to help train a facial recognition system. Although the photos in question were shared under a Creative Commons license, many users say they never imagined their images would be used in this way. Furthermore, the people shown in the images didn't consent to anything. From a report: "This is the dirty little secret of AI training sets. Researchers often just grab whatever images are available in the wild," said NYU School of Law professor Jason Schultz. The latest company to enter this territory was IBM, which in January released a collection of nearly a million photos that were taken from the photo hosting site Flickr and coded to describe the subjects' appearance. IBM promoted the collection to researchers as a progressive step toward reducing bias in facial recognition. But some of the photographers whose images were included in IBM's dataset were surprised and disconcerted when NBC News told them that their photographs had been annotated with details including facial geometry and skin tone and may be used to develop facial recognition algorithms. (NBC News obtained IBM's dataset from a source after the company declined to share it, saying it could be used only by academic or corporate research groups.)
"None of the people I photographed had any idea their images were being used in this way," said Greg Peverill-Conti, a Boston-based public relations executive who has more than 700 photos in IBM's collection, known as a "training dataset." "It seems a little sketchy that IBM can use these pictures without saying anything to anybody," he said. John Smith, who oversees AI research at IBM, said that the company was committed to "protecting the privacy of individuals" and "will work with anyone who requests a URL to be removed from the dataset." Despite IBM's assurances that Flickr users can opt out of the database, NBC News discovered that it's almost impossible to get photos removed. IBM requires photographers to email links to photos they want removed, but the company has not publicly shared the list of Flickr users and photos included in the dataset, so there is no easy way of finding out whose photos are included. IBM did not respond to questions about this process.
"None of the people I photographed had any idea their images were being used in this way," said Greg Peverill-Conti, a Boston-based public relations executive who has more than 700 photos in IBM's collection, known as a "training dataset." "It seems a little sketchy that IBM can use these pictures without saying anything to anybody," he said. John Smith, who oversees AI research at IBM, said that the company was committed to "protecting the privacy of individuals" and "will work with anyone who requests a URL to be removed from the dataset." Despite IBM's assurances that Flickr users can opt out of the database, NBC News discovered that it's almost impossible to get photos removed. IBM requires photographers to email links to photos they want removed, but the company has not publicly shared the list of Flickr users and photos included in the dataset, so there is no easy way of finding out whose photos are included. IBM did not respond to questions about this process.
There's no implication IBM did anything wrong. This is what the Creative Commons licenses are for. What's the story?
Your photos are public. What the hell do you expect?
Someone tell these people how search engines work.
People place images on the public internet, available to world+dog, and then express surprise and dismay that world+dog has access to the images? What's next, shock and dismay upon learning that Zuckerberg knows more about them than the NSA does?
Whether you love or hate the harvard comma, it is generally agreed you don't use it on two-item lists.
Someone had to do it.
Flickr photo sets have been used for computational work loads and data mining for well over a decade, this is hardly NEWs.
https://www.ted.com/talks/blai...
Maybe they could use my image to train sexbot AI's...
Then I would just tell them that I am Captain James T Kirk, and their synthetic panties all drop for me...
An incel can dream
It's pictures available for public conniption ("conniption" was an autocorrect error too funny to correct).
Consumption is just what model training is doing; they are not republishing the pictures in any way, just using them to train models - which do not contain any element of images they train from.
If you put your image in public, how can you be aghast someone has viewed it?
"There is more worth loving than we have strength to love." - Brian Jay Stanley
Although the photos in question were shared under a Creative Commons license, many users say they never imagined their images would be used in this way
Just because you lacked the creativity to consider what was possible with your data doesn't mean there is anything improper has happened when they do use it in such a way.
Also, if you have given away your data thinking that somehow corporations would respect you then you don't really understand what drives corporations.
The reality is that if it's profitable then a corporation will do it. It doesn't matter if it's morally repugnant, illegal or downright evil because if it's possible to make a profit then there will be a corporation that will do it. Note that being illegal typically means they will be fined which they consider a business expense.
Anons need not reply. Questions end with a question mark.
Although the photos in question were shared under a Creative Commons license, many users say they never imagined their images would be used in this way.
Since when is licensing about what you "imagine"?
Funny that I've not yet heard about the CC-as-foreseen license, which apparently billions of people have been using, in earnest, all along.
Copyright governs your ability to distribute copies of other people's work. There's no distribution going on here, so permission of the copyright holder (photographer) was not needed.
It might be governed by personality rights - your right to control how your image is used. You could argue the model's consent is needed before using their facial geometry. But personality rights are generally concerned with control over how others perceive your image. Since there's no public perception or exploitation here, it would be an uphill argument.
AFAIK, there is no basis for prohibiting people from using things you make publicly available (your face every time you walk out in public, unless you wear a burka) to train computer algorithms. Photographers and the press have worked pretty hard to enshrine their right to record images of people in public places. If we want there to be restrictions of using images of people in public places, it'll need to be a new law.
Many of us are not narcissistic attention seeking whores and do not want to be in the spotlight for any reason.
(Not sure why you were down-modded?)
If that is true, you wouldn't have any photos up they could use to train right?
Also any photos used for training, are never in the spotlight as it were.
"There is more worth loving than we have strength to love." - Brian Jay Stanley
Let me guess, the whole quote should he been something like:
"None of the people I photographed had any idea their images were being used in this way, but it's all because I decided to put it on the internet with a licence that allows anyone to do anything with it, without explaining to them what I was going to do."
Sir Tim Berners Lee's new proposal called Solid [https://solid.inrupt.com/] would be a prime choice for managing the access to your data that companies like IBM have for their scanning. Fine grained access to share and audit data in a decentralised distributed network using blockchain.
When shit hits the fan get some of these https://youtu.be/pY-GncsZ-UE
I could care less. Most people photographed also do not care. I can honestly say that if you get offended by a machine learning algorithm looking at your photo, you might have deep psychological issues. It's not a human. No judgement was made.
“It seems a little sketchy that IBM can use these pictures without saying anything to anybody,” he said.
It seems a little sketchy than this photographer didn't explain to the subjects that he was going to post their image online with a licence that allows anyone to do anything with it for any reason.
The fact that they only used creative commons images suggests there's an actual legal issue with proprietary images, but why? If I save an image from a website to my hard drive, without sharing it, does that make me a criminal? I've been training my brain on face recognition with proprietary images for decades. I've even occasionally indirectly made money from the viewing of proprietary images, as has everyone else.
Should I pay a royalty every time I imagine a proprietary image I've previously seen?
This space intentionally left blank
This activity is an express violation of the Privacy rights embedded and explicitly described in both the Canadian and Washington State Constitutions.
Period.
-- Tigger warning: This post may contain tiggers! --
""None of the people I photographed had any idea their images were being used in this way," said Greg Peverill-Conti, a Boston-based public relations executive who has more than 700 photos in IBM's collection, known as a "training dataset." "
Why are you whining? YOU explicitly made that possible. YOU had to elect for each image to be licensed under CC. If the people you photographed are upset by this, they should sue YOU.
Almost all responses here are along the lines of "what did you expect". But it's not that simple.
If I go up to a window in your house and photograph the inside, you don't say "well, I have no problem with that, the windows are transparent after all".
Saying "it's technically possible, so of course someone did it" makes you no better than databrokers like Cambridge Analytica who create psychological profiles based on your Facebook likes and then sell them to, well, anyone really.
Is it technically possible? Yes. Was it something the average user could have anticipated when they pressed the "I agree" button? No.
This is about norms and values. Privacy is a form of "contextual integrity". We have expectation of how much we will get for different situations. People have similar expectations online.
There's no need, because they are not republishing the photos or even legitimately derivative works. What they used from the photos is objective data, not an artistic work built on the photos. You can't copyright objective data, only a particular representation of it. In this case, the photographer's representation is a photo, and IBMs is a neural net.