Meta's Mega AI Model to Power Video Ecosystem

Image credits: Dima Solomin / Unsplash

Meta is powering up its artificial intelligence game.

The tech giant is in the process of developing an AI system that will act as the driving force for Facebook's entire video recommendation engine across all its platforms. This revelation was made by Tom Alison, the head of Facebook, during a recent tech conference in San Francisco.

This grand plan is part of Meta's "technology roadmap that goes to 2026," and it involves creating an AI recommendation model capable of powering both the company's TikTok-like Reels short video service and more traditional, longer videos.

Historically, Meta has used a separate model for each of its products, such as Reels, Groups, and the core Facebook Feed. However, the company is now looking to streamline this process.

Meta's ambitious AI journey involves spending billions on Nvidia graphics processing units (GPUs), the primary chips used by AI researchers for training large language models. These models are used to power popular AI tools like OpenAI's ChatGPT chatbot.

The first phase of Meta's tech roadmap involved transitioning the company's current recommendation systems to GPUs from traditional computer chips, leading to improved overall performance of products.

Meta's interest in large language models was piqued by their ability to handle vast amounts of data and perform general-purpose activities like chatting. This led to the development of a giant recommendation model that could be used across products.

This new "model architecture" was tested on Reels and resulted in an 8% to 10% increase in Reels watch time on the core Facebook app. This success demonstrated that the model was learning from the data much more efficiently than the previous generation.

Now, Meta is in "phase 3" of its re-architecture of the system, which involves validating the technology and pushing it across multiple products. The goal is to power the entire video ecosystem with this single model and improve the responsiveness of recommendations.

Meta's AI journey doesn't stop there. The company is also considering integrating more sophisticated chatting tools into its core Feed and within Groups. This could allow users to interact with a digital assistant for more information or answers to their questions.

In the words of Alison, "I think we have the opportunity to put generative AI in kind of a multiplayer kind of consumer environment." So, watch this space as Meta continues to push the boundaries of AI!

Made with TRUST_AI - see the Charter: https://www.modelprop.co.uk/trust-ai