Facial Recognition's 'Dirty Little Secret' Is Using Millions of Pictures Without Consent

Alfred Osborne
March 15, 2019

BIG BLUE IBM reportedly used almost a million photos from image sharing site Flickr to train its facial recognition tech, but it did so without user consent.

New York University School of Law professor Jason Schultz said, "this is the dirty little secret of AI training sets".

"None of the people I photographed had any idea their images were being used in this way", one photographer told NBC. "IBM has been committed to building responsible, fair and trusted technologies for more than a century and believes it is critical to strive for fairness and accuracy in facial recognition".

NB News reports that getting photos removed is nearly impossible, however. "Now they are being unwillingly or unknowingly cast in the training of systems that could potentially be used in oppressive ways against their communities".

Flights cancelled at Waterloo airport after grounding
Under global rules, Ethiopians are leading the investigation but France's BEA will conduct black box analysis as an adviser. President Donald Trump , an aviation enthusiast with deep ties to Boeing , said he hoped the suspensions would be short.

Democrats seeking primary opponent for Omar
Omar has come under heavy criticism from both Republicans and Democrats for the use of anti-Semitic slurs. 'I'd be pretty uncomfortable supporting Rep.

DraftKings fantasy golf picks: The Players Championship 2019
Don't worry, though - if we've seen anything about TPC Sawgrass , the course will find a way to equalize the field. Adam Hadwin, however, doesn't like to get too geared up for any one hole on any course.

The only problem? The dataset isn't publicly available, only to researchers, so Flickr users and those featured in their images have no way of really knowing if they're included.

IBM has said that it will remove any images from the set that a photographer or subject wishes. Instead, the images were uploaded to Flickr under a "Creative Commons" license, which allows others to use them without paying licensing fees, sometimes for commercial use. "NBC says that assertion contradicts IBM's admission on the dataset's release that it was in response to research by MIT's Joy Buolamwini, but Smith told Biometric Update that the goal of the dataset is to provide insights on the quality of datasets and questions like "(h) ow can we ensure that face image data is sufficiently diverse?"

NBC talked to several people whose images had appeared in IBM's dataset, including a PR executive who has hundreds of images sitting in the collection.

IBM's dataset drew upon a huge collection of around 100 million Creative Commons-licensed images, referred to as the YFCC-100M dataset and released by Flickr's former owner, Yahoo, for research purposes - there are many CC image databases used for academic research into facial recognition, or fun comparison projects. "You are laundering the IP and privacy rights out of the faces", he says. It added, "Individuals can opt-out of this dataset". Hopefully, the company makes it the opt-out process easier.

Other reports by

Discuss This Article