Meta Wants to Make the World's Fastest Computer


 Meta is building an artificial intelligence supercomputer (AI supercomputer) that will be the fastest in the world.

As the name implies, this supercomputer is designed to train machine learning systems. Meta also claims that its AI Research SuperCluster (RSC) is currently one of the fastest computers in the world for its class, and when completed in mid-2022 it will be the fastest in the world.


"Meta is developing what we believe to be the world's fastest AI supercomputer. We named it RSC for AI Research SuperCluster and it will be completed this year," said Meta CEO Mark Zuckerberg as quoted by The Verge, Tuesday (25/1/2022).



This news shows the importance of research in the AI ​​sector for companies like Meta. Companies such as Microsoft and Nvidia are now launching their own artificial intelligence supercomputers, but keep in mind, these supercomputers are different from ordinary supercomputers.


RSC will be used to train various systems in the Meta business, ranging from content moderation algorithms -- which are used to detect hate speech on Facebook and Instagram -- to training augmented reality (AR) features that will be used by Facebook's AR devices.



It is also confirmed that RSC will be used to design user experiences in the metaverse, as quoted by us from The Verge, Tuesday (25/1/2022).


"RSC will help AI Meta researchers to build new and better AI models, also be able to learn from trillions of existing examples; operate in hundreds of different languages; analyze text, images, and videos; develop new augmented reality tools; and much more, " said Kevin Lee and Shubho Sengupta, Meta engineers, in the official Meta blog post.


"We hope that RSC will help us build completely new AI systems that can, for example, perform real-time voice translation for multiple people at once, each speaking a different language, so they can collaborate seamlessly." in research projects or playing AR games together," explained the two.


The development of RSC started 1.5 years ago, when Meta engineers designed various systems for the computer, such as cooling, power, networking, and wiring, all from scratch.


The current phase one RSC running on a system contains 760 Nvidia GGX A100, which contains a total of 6,080 GPUs. GPUs are commonly used as processors for machine learning because they perform better than CPUs.


Meta claims RSC phase two will be completed before 2022 ends, and at that stage RSC will have 16,000 GPUs that can train AI systems with more than a trillion parameters in data sets that can reach exabytes in size.

Previous Post Next Post

Contact Form