Aiming to enter a new age of artificial intelligence, Meta, previously established as Facebook, is developing what it claims would be the world’s strongest AI supercomputing system. One of the world’s strongest AI supercomputers, it’s called the AI Research SuperCluster (RSC). AI-powered real-time engagements, such as letting “huge groupings of individuals, everyone talking a distinct tongue… effortlessly participate on a scientific study” or “enjoy an AR game simultaneously,” are Meta’s goals with the computing behemoth.
The Goals of The Future
The company’s key goals are voice detection, machine learning, plus neuro-linguistic coding. There can be no way around it: the metaverse has to emerge in this Meta-Facebook news statement, as RSC confesses it will utilize its sophisticated new supercomputer to develop AI-driven apps and solutions that eventually serve the virtual world.
“We’re calling it RSC for AI Research SuperCluster and it’ll be complete later this year. The experiences we’re building for the metaverse require enormous compute power – quintillions of operations per second – and RSC will enable new AI models that can learn from trillions of examples, understand hundreds of languages, and more.”
Considering Meta plans to perform tests with hundreds of GPUs on the supercomputer for months on end, its reliability is critical. In order to support the innovative science and technology the firm claims to be striving for, the user interface must also be respectable. It’s a really impressive construct for a venture that began as a completely remote one, especially given the simplicity of use and reliability (including the apparent computational capability).
For the same reasons as any existing AI supercomputer, Meta is building RSC from the ground up using a high-efficiency network fabric to link numerous GPU computation clusters. An Nvidia Quantum 1600 Gb/s InfiniBand two-level Clos fabric connects RSC’s 760 DGX A100 computers at this stage of construction. When completed, the system will link 16,000 graphics processing units (GPUs). Pure Storage FlashArray, Penguin Computing Altus workstations, plus 10 petabytes of Pure Storage FlashBlade make up its memory layer.
“We’re calling it RSC for AI Research SuperCluster and it’ll be complete later this year. The experiences we’re building for the metaverse require enormous compute power – quintillions of operations per second – and RSC will enable new AI models that can learn from trillions of examples, understand hundreds of languages, and more.”
Would Be Completed by the End of 2022
The supercomputer is expected to be completed by the course of this year, according to Meta’s projections. Until now, the project’s whereabouts have been kept secret; from a technological aspect, Meta is ensuring that all material passes through a security assessment procedure to ensure that it has been appropriately masked before being used.
A primary goal of Meta is to develop self-supervised training or transformer-based systems. Scalability is not a problem with these complicated structures. A single paradigm can deal with a variety of information kinds, including sound, writing, and graphics. To educate systems with more than a trillion variables, the RSC was designed from the ground up.
“Ultimately, the work done with RSC will pave the way toward building technologies for the next major computing platform — the metaverse, where AI-driven applications and products will play an important role.
