Meta Platforms has improved the recommendation algorithm used for suggested videos on Facebook, using artificial intelligence for more accurate selection. This has led to a significant increase in video watch time on the social network.
The recommendation AI was first used in Meta’s Reels service, a competitor to TikTok. It is now integrated into Facebook, and in the future it will be employed across all of Meta’s products where possible. According to Facebook’s head, Tom Alison, the introduction of the new algorithm has increased Facebook video watch times by 8-10%.
“This shows that the new AI model learns from the data much more effectively than the previous generation. It’s a good sign that we’re on the right track,” Alison said at the Morgan Stanley technology conference in San Francisco this week.
Until recently, Meta used different recommendation algorithms in Reels, Facebook groups, and the Facebook feed. However, the success of the AI in Reels prompted a wider distribution. This move is part of Meta’s overall strategy to implement artificial intelligence across all its products. The company is investing billions of dollars in purchasing hundreds of thousands of NVIDIA graphical accelerators, essential for creating large language models.
