Facebook divides content on the platform into 22 “cards,” then AI analyzes user behavior to suggest what they might like to see.
"Billions of users go to Facebook and Instagram every day to share stories, connect with people with similar interests, and discover content. To make each person's experience more personalized, we use AI to decide what content appears in their feeds," Nick Clegg, Meta's President of Global Affairs, wrote on the company's blog in late June.
Clegg believes that with the remarkable progress of AI, people are both excited and scared, so the best way to address these concerns is to be frank and open.
Meta said that most content on Facebook and Instagram is divided into 22 "system tags". Each card contains details about how the AI ranks and recommends content. The algorithm will predict the value of news for each user, thereby deciding what will be prioritized to appear in good positions.
First, AI analyzes each person's behavior, habits, and preferences from the posts they post or reshare.
"We combine factors to predict content that is closest to user interests. Many algorithms are based on behavior, others are based on surveys and user feedback," Clegg said.
Social networks also use AI to filter and remove harmful content to help limit the distribution of low-quality articles.
In the near future, Facebook will allow users to click on a content, review activity history to understand why AI recommended this content to them.
Clegg said Facebook also gives users some tools to choose their own experience with AI.
In each suggested video or article, they can also inform the platform whether they are interested or not interested.
The social network said it has released more than 1,000 AI models, libraries, and data for all users and researchers to maximize algorithm transparency.
The library includes data from public Facebook posts, pages, groups, and events.
The Verge Assessing the explosion of AI in recent times has attracted great attention from regulatory agencies around the world.
Khuong Nha (according to The Verge)
