Meta is building a giant AI model to power its video ecosystem

Meta is developing an artificial intelligence (AI) system that’s designed to power Facebook’s entire video recommendation engine across all its platforms, a senior executive has revealed.

Meta is now creating a model that will be able to power both the company’s TikTok-like Reels short video service and more traditional longer videos.

According to Tom Alison, the head of Facebook, who spoke at Morgan Stanley’s tech conference in San Francisco, Meta has used a separate model for each of its products to date – but this is about to change. The news was first reported by CNBC.

This is an obvious attempt to catch up to TikTok. The rival platform runs on an extremely powerful recommendation algorithm that combines multiple machine-learning techniques to deliver relevant and personalized content to users.

TikTok is extremely popular – so much so that Meta has seen the dominance of apps like Facebook and Instagram erode. As a competitive strategy, Meta has been pushing Reels to ever larger audiences. However, clearly, it has decided that more is needed.

That’s why, as Alison said, Meta has been spending billions of dollars on Nvidia graphics processing units (GPUs). They’ve become the primary chips used by AI researchers to train the large language models used to power generative AI chatbots such as ChatGPT.

The company will switch current recommendation systems to GPUs and hopes to improve the overall performance of its products.

Alison says that Meta has already tested the upcoming system on Reels and saw that it helped the firm obtain an 8-10% gain in Reels watch time, and it’s now time to expand the new technology across multiple products.

“Instead of just powering Reels, we’re working on a project to power our entire video ecosystem with this single model, and then can we add our Feed recommendation product to also be served by this model,” Alison said.

“If we get this right, not only will the recommendations be kind of more engaging and more relevant, but we think the responsiveness of them can improve as well.”

Meta is also looking to enhance its chatbot assistant. In a use case cited by Alison, a user who was recommended content about Taylor Swift could then utilize the AI-powered Meta Assistant to find out when her next tour date is rather than wait for a commenter to answer the question or to search for the date elsewhere on the web.

In another case example involving Groups, a user could rely on Meta AI to answer a baking-related question sooner than another user would.