Meta has unveiled plans for a groundbreaking AI system set to power Facebook’s entire video recommendation engine. Tom Alison, Facebook’s head, revealed at the Morgan Stanley tech conference that this initiative is a crucial part of Meta’s ambitious technology roadmap extending until 2026.
Traditionally, Meta employed separate AI models for its diverse products, including the TikTok-inspired Reels, Groups, and the core Facebook Feed. However, Meta is now investing heavily in a unified recommendation model that spans across all its platforms, promising a seamless and engaging user experience.
To accomplish this feat, Meta has committed substantial resources, channeling billions of dollars into Nvidia graphics processing units (GPUs), widely acknowledged as the primary chips driving advancements in AI research. The shift from conventional computer chips to GPUs marks the first phase of Meta’s technological evolution, enhancing the overall performance of its products.
Driven by the explosive growth in interest surrounding large language models (LLMs), Meta recognised the potential of these models to handle vast amounts of data and diverse tasks. Last year, the company introduced a new model architecture and tested it on Reels, achieving an impressive 8% to 10% gain in watch time on the core Facebook app.
Now entering the third phase of its re-architecture, Meta aims to validate and extend this technology to power its entire video ecosystem. Tom Alison explained that the goal is not only to make recommendations more engaging and relevant but also to improve their responsiveness. If successful, users can expect a more interconnected experience, where content discovered in Reels seamlessly translates to personalised recommendations in the Feed.
Meta has strategically amassed a significant stockpile of GPUs, anticipating their use in broader generative AI efforts. This includes the development of digital assistants, with plans to incorporate sophisticated chatting tools into the core Feed. Users might soon interact with Meta AI, effortlessly seeking more information about recommended content or asking questions within specialised groups.
The company’s vision extends beyond individualised interactions, as Meta explores the integration of generative AI into a multiplayer consumer environment. This innovative approach positions Meta at the forefront of AI development, pushing boundaries to enhance user engagement and interaction across its diverse platforms.
