Meta is planning to create a centralized AI model that will manage its entire video ecosystem – both short-form and long-form.
The first step has already been taken, which was to change Meta’s existing video recommendation system to GPUs.
By doing this, Meta has been able to increase its reel watch time by 8%-10%.
Meta has been taking big leaps in the AI industry – its giant AI model is the newest addition to it. Using this new AI model, it plans to power its entire video ecosystem.
Talking about the same, Tom Allison, the head of Facebook said that this is all a part of the “technology roadmap that goes to 2026”.
This will basically include creating an AI recommendation model that can power both the small reel-like videos as well as the longer videos.
As of now, Meta uses separate models for each product. For example, the model that supports short-form videos is different from the one that supports in-feed content. However, that will all change with the new AI project it is investing in.
The inspiration behind this idea came last year when AI tools became increasingly popular. The executives were literally speechless by the growing capacity of AI i.e. its power to handle millions of data and offer solutions within seconds. That’s when they decided that their video ecosystem needed to leverage AI.
By centralizing video recommendations, they’ll be able to create a more seamless experience for Meta users. For example, if you like a video in the reels section and interact with it, similar types of videos will be shown on your main feed as well, and vice versa.
Read More: Meta to start labeling AI-generated images on Facebook, Instagram, and other platforms
Why This Sudden Change of Plans?
It’s believed that Meta is creating this AI model to catch up with TikTok. TikTok runs on a very powerful algorithm that perfectly recommends the right content to each user based on their behavior. It’s also one of the first apps to introduce short-format videos to the public that gave them an edge.
It’s worth noting that Meta has experimented with other methods, too. For starters, it simply started pushing reels to more people and while that did increase reach, it just didn’t give the engagement Meta was hoping for. So, to catch up, it was clear that it had to up its technology.
The first step in the process was to switch Meta’s current video recommendation system to GPUs. For this, Meta has reportedly spent billions of dollars on Nvidia GPUs. By switching to GPUs, Meta has been able to get 8%-10% more reel watch time.
Read More: Nvidia adds $1 trillion to its market cap in just 3 months this year, beating other tech giants
The next step will be to push this change across all Meta products. By the end of this process, Meta is hoping to increase not only watch time but also engagement and responsiveness from the user’s side.
Meta is also planning to add an AI-powered chatbot. This will help users get answers without waiting for another user to comment. For example, if a clip from a movie is posted, they can find out the name of the movie through the chatbot instead of scrolling through the comments.
While all of this is indeed very exciting, the advancement in AI needs to be taken with a pinch of salt, especially with Meta – a company that was recently accused of running a massive data processing business as well as having a general disregard for user privacy.
>>> Read full article>>>
Copyright for syndicated content belongs to the linked Source : TechReport – https://techreport.com/news/meta-is-investing-in-a-massive-ai-model-that-will-power-its-entire-video-ecosystem/