Meta explains how its AI algorithms work on Instagram and Facebook

For people that are new to the platforms, and have no data to learn from, Meta has developed a "few-shot learning system called Meta Interest Learner". (Image: Meta)

Meta has explained how its AI algorithms decide what you see on your timeline on Instagram and Facebook.

More than 20 percent of the content you see on your feeds is decided by AI, curated from people, organisations or public accounts that you follow.

Also read | Meta adds parental control tools in Instagram and Messenger

For this to work, the AI needs to understand the content it's recommending. Meta says that over the years, they have poured a lot of research into this area and this helps the AI with recognising visuals, objects, text and audio.

The AI can classify posts according to topic/genre, can predict hashtags, and find similar posts and clusters. Meta's retrieval systems for content take just, "hundredths of a second to narrow billions of pieces of content down to thousands and then to a few hundred that are relevant to a particular person’s interests".

The ranking system then takes over and classifies these posts based on, "pointwise and listwise predictions". Meta says their AI algorithms have contributed to a 15 percent increase in watch time on the Reels video player on Facebook.

Also read | Meta partners CAIT to upskill one million traders on WhatsApp Business app

For people that are new to the platforms, and have no data to learn from, Meta has developed a "few-shot learning system called Meta Interest Learner". This helps the platform match posts to prospective audiences based on general interests.

To learn how people choose and interact with posts, Meta uses cutting-edge embedding learning and graph learning methods. This helps control the delivery frequency of posts on a timeline and helps in content discovery by pushing posts according to personalised interest.