Meta is building the fastest supercomputer in the world (to plug us into the Metaverse)
From the moment Facebook rebranded itself to Meta the online space exploded with some pretty polarized opinions on the matter. The plan behind this move isn’t crystal clear at the moment but it seems that Facebook wants to create a 3D, virtual version of the Internet (capital letter needed).
Now, Facebook has just shy of 3 billion monthly active users, and it’d take an enormous amount of computing power to plug them all in a virtual reality environment. The news that the company is attempting to build the fastest supercomputer in the world ties well with the whole idea.
To the news in question - Meta has completed phase one of its new AI supercomputer project. The name of the machine is AI Research SuperCluster (RSC) and when it’s completely finished later this year, Meta believes it will be the fastest supercomputer on the planet.
This new RSC machine will be able to perform floating point calculations with 5 exaflops (approximately 1,000,000,000 times faster than a regular gaming PC). Meta wants to use all this power to develop AI algorithms that can learn from trillions of examples.
Another potential use of this supercomputer is building better augmented reality tools, as the machine will be able to “seamlessly analyze text, images and video together." According to Meta, this will be a key pillar in the vision of the company for its Metaverse project.
The RSC will also be used to find hate speech and harmful content (most likely in Facebook and Instagram), and root it out. The AI can be trained to do this with great efficiency, according to Meta.
The first stage of the RSC project has now been completed and the machine is operational. At its current stage the supercomputer has 760 Nvidia DGX A100 systems with a total of 6,080 graphic processing units.
This setup is already faster than many supercomputers out there but Meta plans to increase the number of GPUs in RSC to 16,000. That next step will boost the performance of the system by around 2.5 times, and allow the computer to work on chunks with exabyte size (1 exabyte is equivalent to 36,000 years of high-quality video).
"We expect such a step function change in compute capability to enable us not only to create more accurate AI models for our existing services, but also to enable completely new user experiences, especially in the metaverse," Lee and Sengupta wrote.
The rebranding of Facebook to Meta was officially introduced at Facebook Connect on October 28, 2021. Zuckerberg presented the change with a very futuristic connotation.
“In this future, you will be able to teleport instantly as a hologram to be at the office without a commute, at a concert with friends, or in your parents’ living room to catch up,” Zuckerberg wrote in a letter.
It is possible, though, that Facebook wanted a fresh start after numerous security breaches, whistleblowers, and leaks. The company has been a subject of scrutiny ever since the Cambridge Analytica case came to light.
At the moment, there’s not much behind this rebranding - Facebook, Instagram and WhatsApp all operate the way they used to, and the parent company doesn’t seem to be changing structurally all that much.
It remains to be seen if the Metaverse plan will come to fruition, and whether we’re going to experience the blockbuster movie The Matrix first hand.
Now, Facebook has just shy of 3 billion monthly active users, and it’d take an enormous amount of computing power to plug them all in a virtual reality environment. The news that the company is attempting to build the fastest supercomputer in the world ties well with the whole idea.
What can the RSC do?
Phase one is done and the machine is fully operational
This new RSC machine will be able to perform floating point calculations with 5 exaflops (approximately 1,000,000,000 times faster than a regular gaming PC). Meta wants to use all this power to develop AI algorithms that can learn from trillions of examples.
Another potential use of this supercomputer is building better augmented reality tools, as the machine will be able to “seamlessly analyze text, images and video together." According to Meta, this will be a key pillar in the vision of the company for its Metaverse project.
"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," technical program manager Kevin Lee and software engineer Shubho Sengupta wrote in a blog post.
What’s next?
The first stage of the RSC project has now been completed and the machine is operational. At its current stage the supercomputer has 760 Nvidia DGX A100 systems with a total of 6,080 graphic processing units.
This setup is already faster than many supercomputers out there but Meta plans to increase the number of GPUs in RSC to 16,000. That next step will boost the performance of the system by around 2.5 times, and allow the computer to work on chunks with exabyte size (1 exabyte is equivalent to 36,000 years of high-quality video).
"We expect such a step function change in compute capability to enable us not only to create more accurate AI models for our existing services, but also to enable completely new user experiences, especially in the metaverse," Lee and Sengupta wrote.
What is Meta?
Could this be our future?
The rebranding of Facebook to Meta was officially introduced at Facebook Connect on October 28, 2021. Zuckerberg presented the change with a very futuristic connotation.
It is possible, though, that Facebook wanted a fresh start after numerous security breaches, whistleblowers, and leaks. The company has been a subject of scrutiny ever since the Cambridge Analytica case came to light.
At the moment, there’s not much behind this rebranding - Facebook, Instagram and WhatsApp all operate the way they used to, and the parent company doesn’t seem to be changing structurally all that much.
You may also find interesting:
Things that are NOT allowed: