Is a Mark Zuckerberg-Backed AI Firm Scraping Your Personal Data? — 'Taskers' Speak Out

An unseen workforce has been feeding information to machines that allows artificial intelligence to train on online material. Recently, The Guardian looked into the claims of taskers who had been working for a Mark Zuckerberg-backed artificial intelligence (AI) company.

They claimed that they had been promised they would not be exposed to certain kinds of information, but were made to tag and catalog sensitive material as part of their job anyway.

Scale AI, in which META has a 49 percent stake, uses a platform called Outlier to carry out the tasks. The company has taken in experts from multiple fields to exploit their knowledge to feed data into artificial intelligence models across multiple areas of study. However, workers have said that they are being made to scrap the personal data of individuals from social media, which seems to have nothing to do with training high-level systems.

Taskers are the human labor that keeps AI models fed with information that can allow them to generate data and images. This entails scraping information from the internet, which includes, but is not limited to, copyrighted material that can be found online.

Everything from social media accounts to art and literature that has been posted online is up for grabs for these machines to be trained. Workers have claimed to tag people’s photos by name, and even establish their relationships with other people in the photo.

Speaking to The Guardian, a contractor stated, "I don't think people quite understood that there'd be somebody on a desk in a random state, looking at your [social media] profile, using it to generate AI data."

However, most of these taskers are desperate to earn a living, a situation that has been made worse in many industries with the very machines that the people end up training. Another contractor, while speaking to news outlet, noted, "A lot of us were really desperate. Many people really needed this job, myself included, and really tried to make the best of a bad situation."

Among the many tasks handed to them to train the machine, some of the contractors suggested that they were asked to label users’ ages across multiple pictures and arrange them on the basis of age. A Scale AI source that spoke to The Guardian, however, was not aware of any such task. The source also claimed that accounts set to private are not scrapped.

Taskers have stated that they were asked to use copyrighted work to train the AI that would make its own artistic works. The source, on the other hand, said that there was no such mandate, and in fact, taskers were specifically asked not to use copyrighted work. However, taskers suggested that they were only given instructions not to use AI images to train the system.