Facebook Still Struggling With Privacy Issues - Is There Light In The Tunnel?

source: Flickr

Facebook privacy issues continue. As Reuters first reported, the social media giant has been using a team of 260 contract workers in Hyderabad, India, to label status updates and photos from 2014 onward with pre-defined keywords. Reportedly, the aim was to create training data for the AI systems.

Using machine learning and AI to sort content on platforms is not new to many high tech companies, including Facebook. However,  to do its job, the software must be trained so that it could differentiate between the diverse types of content. 

Therefore, the human touch is needed to categorize and label some sample data first. The process is widely known as ''data annotation." What Facebook did was not an exception. Many fellow tech companies use data annotation to help AI identify traffic lights or pedestrians in videos.

Regular internet users like you and me have also performed this type of work without even realizing it. Remember Google's CAPTCHA system, which requires you to prove that you are a human and makes you identify objects? This information is also used for AI training purposes.

As Reuters pointed out, Facebook worked along with the Hyderabad-based firm Wipro Ltd. on its data annotation project. Wipro hired 260 contract workers to go through Facebook and Instagram content and annotate posts according to five pre-defined categories. 

The categories include the content of the post ( is it a selfie or a picture of a dog, food, etc.); the occasion ( is it for a birthday for instance); and the author's intent ( was it a joke or a source of inspiration). Each post had to be verified by two workers to ensure accuracy, anonymous Wipro sources told Reuters, highlighting that they had to go through 700 items each day.

The anonymous Wipro employee also revealed they could see everything from text-based status updates to video streaming and images, including these shared privately. 

Facebook did not decline the news saying that posts and images for the analysis were random and did not include any details of who was their author. However, the social media firm admitted that posts with comments and screenshots could reveal the usernames.

Facebook told Reuters that its internal legal team approved all data-labeling efforts. In addition to that, the company also established an auditing system to make sure that the data annotation work meets the privacy expectations of the users.

Legal experts say that such a practice, although well-motivated, might be illegal if it includes content posted by European Union citizens as it might not be in line with the GDPR which set strict limits on how companies can collect and use personal data.

What do you think? Do you believe that '' the future is private'' as Mark Zuckerberg stated earlier last week?