Zuckerberg: Meta owns the largest supercomputer for artificial intelligence research

Meta announced through a post by Mark Zuckerberg – the company’s CEO – that it has built an “artificial intelligence supercomputer”, a very fast computer designed to train and enhance machine learning systems, according to his publication on Monday.

“META has developed what we believe is the world’s fastest artificial intelligence supercomputer… We call it RSC (RSC), which stands for AI Research Super Cluster and will be completed later this year,” Zuckerberg said in his letter.

AI research has become central to giants like META, as its competitors, including Nvidia and Microsoft, have published ambitious designs for their “supercomputer” systems, which are a variant of the concept. traditional supercomputers.

RSC will have a wide range of applications it is working on, which will include augmented reality applications that will eventually appear in the company’s AR devices, and even content editing algorithms that detect “hate speech” on Instagram. (Instagram) and Facebook (Facebook).

Meta will use RSC to develop new services on metaverse, the virtual world that the company intends to develop.

“RSC will help Meta AI researchers build new AI models that can learn from trillions of examples, work across hundreds of different languages, seamlessly analyze text, images, and video together, develop new augmented reality tools, and much more,” the company said in a blog post. “.

“We hope that RSC will help us build completely new AI systems, for example, it can play voice translations for large groups of people, each speaking a different language, so they can seamlessly collaborate on a research project or play a computer game together.” .

This project began a year and a half ago, when META engineers began working on the various systems that support AI. This means designing power, cabling, networking and cooling systems.

The first phase of the RSC is already complete, comprising 760 systems, and consisting of 6,080 graphics processing units (GPUs) connected.

RSC from Meta will be used to develop new services on Metaverse (Reuters)

This setting has already improved performance compared to traditional computer searches by 20 times, according to the company.

But the biggest ambition is for the company to complete before the end of 2022 the second phase of the RSC. It will then have approximately 16,000 total GPUs, which will enable it to train AI systems.

This may seem like a bunch of numbers, but the real benefit of supercomputers lies in their applications, not in theoretical speed or peak performance.

Meta will start creating new systems, trying to improve their algorithms to reconcile their users. And if this happens, META could become a fearsome force for competitors such as Apple, Google and Microsoft.

Leave a Reply

Your email address will not be published.