2 min read
In another shocking news about Facebook’s illicit activities regarding user privacy, a new report from Reuters reveals that Facebook’s contract workers have been looking at private posts of users on Facebook and Instagram to train their AI systems.
In a day and age of ‘Big Data’, social media corporations from around the world use machine learning and AI tools to sort content on their massive platforms. However, in order to do that these tools or systems need to be trained by analyzing a huge amount of data so that they can identify different types of content. This process is called ‘Data annotation’ where algorithms analyze huge chunks of data which is later categorized by humans.
“It’s a core part of what you need,” Facebook’s Nipun Mathur, director of product management for AI, told Reuters.
The report by Reuters focused on an Indian firm, ‘WiPro’. The Indian company employed over 260 people to annotate data according to three categories:
- Content of the post ( e.g if is it a picture of person or food)
- Occasion (e.g a birthday party or wedding day)
- Author’s intent (e.g are they making a joke or trying to inspire someone)
The employees sort a wide range of content on Facebook and Instagram which includes Stories, photos, shared links, videos and status updates. The employees annotate this data for roughly 700 times a day.
Interestingly, such data annotation projects have become quite common and are regularly outsourced to countries where human labor is cheaper. For example, in China, there are a lot of offices which train self-driving cars by providing the required data. However, this type of work becomes dangerous when the data required is the personal information of users and is taken without their permission or consent.
This may be concerning for Facebook since this content labeling program may raise new privacy issues for Facebook considering the social media giant is already facing regulatory investigations worldwide.
