The machine will help the social media giant to create new services for the virtual world and to better moderate contents on its platforms.
Meta (FB) – Get Meta Platforms Inc. Class A Report formerly Facebook, has no intention of letting Microsoft (MSFT) – Get Microsoft Corporation Report, Samsung (SSNLF) or Nvidia (NVDA) – Get NVIDIA Corporation Report take the lead in the race to mold and shape the metaverse, a virtual reality construct intended to supplant the internet as we know it today.
The social media giant has just unveiled a machine, which confirms its ambitions for this world of which it has made its future sources of growth and which represents its best marketing campaign yet to make people forget its scandals.
The company has built Research SuperCluster, or RSC, a supercomputer that is already up and running but is still being built. It will allow Meta to use data from its various platforms, especially the Facebook social network, to train AI engines, particularly in the field of language processing or ‘picture.
RSC will be, claims Meta on a Blog post, the fastest artificial intelligence supercomputer in the world when it’s fully built out in mid-2022, ahead of the Japanese supercomputer Fugaku, which has been at the top of the supercomputer rankings of the independent site top500.org since 2020.
It’s worth noting that Meta is defining the power of its computer differently from how conventional and more technically powerful supercomputers are measured.
Supercomputers are extremely fast and powerful machines built to do complex calculations not possible with a regular home computer.
Twenty Times faster
The company plans to use RSC to build technologies and to create new services for the metaverse, the new buzzword in the tech world.
“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 Kevin Lee, Meta’s technical program manager, and Shubho Sengupta, a company software engineer.
“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,” Lee and Sengupta added.
The company said its supercomputer will incorporate “real-world examples” from its own systems into training its AI. And that its previous efforts used only open-source and other publicly available data sets
Artificial intelligence engines are driven by ingesting massive amounts of data. For example, text for language processing models, or video images for machine vision models.
RSC has already started training large natural language processing and computer vision models. This will be used to support content moderation on Meta’s platforms, an extremely sensitive subject for the company.
For example, it has created real-time voice translation systems of ever more languages, dialects, accents, or models able to study long videos taking into account more data, to develop a voice recognition model able to work even in complex situations such as a party or a concert.
On the technical side, RCS has, for its Phase 1, 760 Nvidia DGX A100 computers representing 6,080 GPUs, which already allows it during the first tests to work 20 times faster than its predecessor on computer vision and run systems of natural language understanding at scale three times faster.
To train models with tens of billions of parameters, it takes three weeks instead of nine.
