Meta is using AI to make Instagram (even) more addictive

A Meta recently presented its financial results for the second quarter and the company's co-founder and CEO, Mark Zuckerberg, took the opportunity to share some issues relating to Instagram growth.

According to Zuckerberg, time spent on Instagram by users had an increase in the order of two digits over the same period last year.

The leader of Meta attributed this growth to the changes made to the recommendation algorithm – both in feed and in Reels videos – and which result from the company's bet on Artificial Intelligence.

How does the business Insider site, Zuckerberg explained that the recommendations of feed and Reels are currently trained with a large language model, which is “fed” with all publications and videos publicly available on Instagram so that you have information not only about topics but also the most appropriate tone.

“ Our recommendations [of Instagram] are also becoming more personalized, presenting more recent content while giving people more direct control over what they see”, said Zuckerberg, explaining that Instagram Reels combine “a faster inference with a new architecture that uses a more detailed user history to improve predictions”.

According to Zuckerberg, this system made it possible for “an increase of 15 basis points in Instagram sessions”, highlighting still “sharing and time spent” by users on the social network.

This means that Meta is developing Artificial Intelligence models for more than launching functionalities on its social networks, taking advantage of the investment of many billions of dollars that it has made in this technology to change (fundamentally) how content is recommended on Instagram.

Thus, the idea passes not only to allow content creators to reach more people with this system of recommendations, but also to keep Instagram users longer connected to the app and increase interaction with publications.