Meta plans to shape the metaverse with massive AI computers

A behemoth called the Research SuperCluster, or RSC, which Meta. Previously known as Facebook, has built, is one of the world’s fastest supercomputers. CEO Mark Zuckerberg claims it is the fastest machine ever built. To handle AI tasks with 6,080 graphics processing units packaged across 760 Nvidia A100 modules.

In terms of processing power, this supercomputer is comparable to the Perlmutter supercomputer. Which uses over 6,000 of the same Nvidia GPUs, currently ranked as the fifth-fastest supercomputer in the world. In the second phase, Meta plans to increase performance by 2.5 with the addition of 16,000 GPUs this year.

A number of Meta’s research projects will use RSC. Including “multimodal” AI based on the confluence of sound, images, and actions as opposed to just one type of input data. Such information might be helpful for dealing with one of Facebook’s biggest issues, spotting harmful content.

RSC is a major investment by Meta, which is a leading AI company. That aims to use it to create the company’s newest priority: a virtual realm called the metaverse. The speech recognition system could be powerful enough to translate speech for a large group of people who speak different languages simultaneously.

Meta bragging about its massive computer

Mark Zuckerberg, CEO of Meta, said that the experiences for the metaverse require an enormous amount of computing power. “RSC will enable AI models that can learn from trillions of examples, understand hundreds of languages, and more.”

A Google blog post states that RSC is about 20 times faster than Nvidia’s earlier 2017-era machine at training an AI to identify objects in photos, according to a blog post by Meta researchers Kevin Lee and Shubho Sengupta. The decoding of human speech is three times faster than that of machine speech.

Normally, when we talk about artificial intelligence, we’re referring to a type of processing data similar to the way our brains do, called machine learning. It is a revolutionary technology because AI models are trained by working with real-world data. A computer program using Artificial Intelligence can learn what cat faces look like based on thousands of cat photographs, rather than trying to describe the entirety of feline fur, whiskers, eyes, and ears using traditional programming.

Leave a Comment