Training large models requires high-performance computing infrastructures, and the AI research team at Meta has been making efforts on this for years.
Social media giant Meta Platforms (NASDAQ: FB) has announced that it has built an artificial intelligence (AI) supercomputer. According to Meta founder Mark Zuckerberg, the AI supercomputer, AI Research SuperCluster (RSC), is one of the fastest in the world. The latest development is part of Meta’s plans to build a virtual metaverse. Formerly Facebook, Meta changed its name in October, re-introducing itself as a “social technology company.” After adopting the new name, based on the sci-fi term “metaverse,” Meta has been working on incorporating the virtual word under its brand. Zuckerberg noted that the new AI RSC would be the fastest machine in the world after its completion in mid-2022.
Meta Announces AI Supercomputer
While announcing the AI supercomputer, Meta said in a blog post that developing the coming generation of advanced AI is required. The company wrote that advanced AI needs new computers that can handle quintillions of operations per second. Hence, the introduction of the AI supercomputer by Meta.
Already, Meta researchers have begun using RSC to train large models in natural language processing (NLP) for its AI supercomputer. In the press release, Meta said RSC would help its AI researchers to build new and better AI models. The company hopes that the new AI models will work with different languages, easily deciphering texts, images, and videos.
“Our researchers will be able to train the largest models needed to develop advanced AI for computer vision, NLP, speech recognition, and more. 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.”
Eventually, the result RSC produces will be used to build technologies for the metaverse, which is the next computing platform.
Training large models requires high-performance computing infrastructures, and the AI research team at Meta has been making efforts on this for years. The first generation was designed in 2017, and it has 22,000 NVIDIA V100 Tensor Core GPUs in a single cluster that performs 35,000 training jobs per day. Since 2017, Meta researchers have used the infrastructure as a standard in reliability, performance, and productivity.
NVIDIA, Others, Support Meta’s AI RSC Mission
Furthermore, Meta explained that it could develop its AI supercomputer with support from some partners. These same partners also helped the company in developing ts first generation of AI infrastructure in 2017. The partners included Penguin Computing, which collaborated with Meta’s operation teams on hardware integration. Pure Storage also worked with Meta in providing robust and scalable storage solutions to the course. In addition, Meta said technology company Nvidia (NASDAQ: NVDA) provided its AI computing technologies for the development of the AI supercomputer.
At the time of writing, Meta Platforms stock is down 1.89% in premarket trading as it trades at $302.88. Except for an 1104% increase over the past year, Meta has been consistently plunging.
Ibukun is a crypto/finance writer interested in passing relevant information, using non-complex words to reach all kinds of audience. Apart from writing, she likes to see movies, cook, and explore restaurants in the city of Lagos, where she resides.