Meta Building Giant AI Model to Power Entire Video Ecosystem

Facebook chief Tom Alison says parent company Meta Platforms is building a giant AI model that will eventually “power our entire video ecosystem.” Speaking at the Morgan Stanley Technology, Media & Telecom Conference this week, Alison said the model will drive the company’s video recommendation engine across all platforms that host long-form video as well as the short-form Reels, which are limited to 90 seconds. Alison said the company began experimenting with the new, super-sized AI model last year and found that it helped improve Facebook’s Reels watch time by anywhere from 8-10 percent.

Alison said the results indicated “the model was ‘learning from the data much more efficiently than the previous generation,’” according to CNBC, which said the executive described the single model plan as “part of Meta’s ‘technology roadmap that goes to 2026.’”

“Alison said Meta is working to continue validating the use of large language models (LLMs) and generative AI to power more products, and scale it,” Quartz writes.

At this point, “we’ve really focused on kind of investing more in making sure that we can scale these models up with the right kind of hardware,” Alison said, per the CNBC report.

Meta Platforms CEO Mark Zuckerberg discussed the roadmap and “massive compute infrastructure” earlier this year in a Threads Reel.

“To date, Meta has typically used a separate model for each of its products, such as Reels, Groups and the core Facebook Feed, Alison said onstage … in San Francisco,” per CNBC.

Alison shared that Facebook Reels represents roughly 33 percent third of Facebook users’ video viewing time, and that “video time on the platform is worth over 50 percent of overall time spent on the site,” Quartz says.

“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 on Engadget, adding that “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.”

No Comments Yet

You can be the first to comment!

Leave a comment

You must be logged in to post a comment.