[ad_1]
A new report from Reuters reveals that contractors are consulting private publications on Facebook and Instagram in order to label them for artificial intelligence systems.
Like many tech companies, Facebook uses machine learning and AI to sort the content on its platforms. But to do this, the software must be trained to identify different types of content. To form these algorithms, they must analyze sample data, all of which must be classified and tagged by humans – a process called "annotation of data."
Reuters'Report focuses on the Indian outsourcing company WiPro, which employed up to 260 workers to annotate positions according to five categories. These include the content of the publication (whether it's a selfie, for example, or a food picture); on the occasion (is it for a birthday or a wedding); and the intention of the author (do they make a joke, try to inspire other people or organize a party)?
WiPro employees need to sort through a range of Facebook and Instagram content, including status updates, videos, photos, shared links and stories. Each worker checks the accuracy of each piece of content, which annotates about 700 articles per day.
Facebook confirmed to Reuters that the content reviewed by WiPro employees includes private publications shared by a limited number of friends, and that the data sometimes include user names and other sensitive information. Facebook claims to have 200 content labeling projects in the world, employing thousands of people.
"It's an essential part of what you need," said Nipun Mathur, director of product management for AI at Facebook. Reuters. "I do not see the need to go."
These data annotation projects are essential to the development of artificial intelligence and have become somewhat like call centers, outsourced to countries where human labor costs less.
In China, for example, huge offices of people tag images of autonomous cars to train them in the identification of cyclists and pedestrians. Most people have done this kind of work without even knowing it. The Google CAPTCHA system, which asks you to identify objects in images to "prove" that you are human, is used to scan information and train the AI.
This type of work is necessary, but disturbing when the data in question are private. Recent surveys have highlighted how work teams tag sensitive information collected by Amazon Echo devices and Ring security cameras. When you talk to Alexa, you do not imagine that anyone else is listening to your conversation, but that's exactly what can happen.
The problem is even more troubling when the work is outsourced to companies that may have less stringent security and privacy standards than larger technology companies.
Facebook states that its legal and privacy teams are approving all efforts to tag the data, and the company said Reuters that he has recently put in place an audit system "to ensure that privacy expectations are met and that the settings in place are working as intended".
However, the company could still violate the recent EU RPP rules, which set strict limits on how companies can collect and use their personal data.
Facebook says that data labeled by human workers is used to form a number of machine learning systems. These include the content recommendation in the company's shopping function; describe photos and videos for visually impaired users; and sort the messages so that certain advertisements do not appear next to political or adult content.
[ad_2]
Source link