Image credits: Meta
/
News Article

Meta to Complete Fastest AI Supercomputer

By Cinthya Alaniz Salazar | Tue, 01/25/2022 - 15:18

After two years of development, the world’s fastest AI supercomputer will be built by mid-summer 2022, announced Meta. The supercomputer is an essential building block of the company’s Metaverse business vision.

“The experiences we are building for the metaverse require enormous compute power (quintillions of operations / second!) and RSC will enable new AI models that can learn from trillions of examples, understand hundreds of languages and more,” Mark Zuckerberg shared on Facebook.

By July, Meta aims to have tripled the capacity of the supercomputer’s data center, AI Research SuperCluster (RSC), by 16,000 central processing units (CPUs), according to Kevin Lee, Technical Program Manager, Meta. Building the processing capability of RSC will allow for the development of more accurate AI models that can learn from trillions of data samples so that ultimately the supercomputer is capable of working across different languages, simultaneously analyze text, images and video and inform the development of augmented reality tools.

Initial applications, however, will likely be limited to the filtering and identification of harmful content until the system is fed enough datasets required to build new AI systems such as real-time voice translation. A relevant feature to augmented reality gaming experiences, which represent one of the most outright market opportunities of the Metaverse. Intrinsically, RSC will play a leading role in the identification of new market avenues for the company to develop in parallel to infrastructure of the Metaverse.

“We hope RSC will help us build entirely new AI systems that can, for example, power real-time voice translations to large groups of people, each speaking a different language, so they can seamlessly collaborate on a research project or play an AR game together,” wrote Meta’s researchers. “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.”

Critics have already raised concerns about the project's development such as where the data sets that will be fed to the supercomputer are being sourced from, insinuating that they will likely come from Meta’s over 2.8 billion daily users. Nevertheless, the company press release assures that before data is imported to RSC it will go through a privacy review process “to confirm it has been correctly anonymized,” but did not share details about where the data sets were sourced from. Moreover, it also added that the entire data path, from storage systems to the GPUs, will be encrypted end-to-end even before they are used to train AI models. Given the company’s history, the project will likely be under close inspection from regulatory authorities.

Regardless of public sentiment, which ranges from excitement to concern, Meta will continue building on the capacity of this supercomputer even after Phase 2 of the project is complete. Meta has ambitions to continue building on the supercomputer’s processing capacity throughout the remainder of the year and increase AI performance.

The data used in this article was sourced from:  
Mark Zuckerberg, Meta
Photo by:   Meta
Cinthya Alaniz Salazar Cinthya Alaniz Salazar Journalist & Industry Analyst