Meta is advancing its AI capabilities, aiming to enhance Facebook’s video recommendation system across its platforms, announced Tom Alison, Facebook’s head, at a tech conference in San Francisco. This initiative is part of a broader “technology roadmap to 2026,” which includes integrating a unified AI model for both short and traditional video content. Traditionally employing separate models for its diverse services, Meta’s new approach involves consolidating these into a single, more efficient system.
The company’s shift to NVIDIA’s GPUs — a move driven by the chips’ effectiveness in training complex AI models — marks a significant investment in AI technology. This transition to GPUs is already improving product performance, as evidenced by an increase in Reels watch time on Facebook.
Currently, Meta is testing this AI model architecture across various products, aiming for a more cohesive user experience. For example, users could see more related content in their Feed based on their interactions with Reels. This effort represents just one part of Meta’s broader generative AI strategy, which also explores digital assistants and enhanced chatting features within Facebook’s ecosystem.
Alison mentioned that they have significantly concentrated on investing in the scalability of these models through the appropriate hardware. He elaborated on their efforts beyond just enhancing Reels, stating they are developing a project aimed at utilizing this singular model to govern their entire video ecosystem, with plans to incorporate their Feed recommendation product into this model’s capabilities. Alison expressed optimism that achieving this would not only make the recommendations more engaging and relevant but also enhance their responsiveness. He also shared his belief in the potential of integrating generative AI into a multiplayer consumer environment.
Featured image: Credit: Meta
