Meta builds an AI supercomputer
| Business Top stories | Yahoo news
Meta on Monday announced its new AI Research SuperCluster (RSC), a supercomputer intended to be used for AI research projects. The supercomputer, which has been in the works for two years, could eventually help Meta develop much more powerful AI software, potentially useful for tricky tasks like detecting hate speech in messages.
“With RSC, you can train faster [AI] models that use multimodal cues – for example, to understand the context of a message, including language, images, and tone,” Shubho Sengupta, software engineer at Meta (Facebook)told CNN Business.
Supercomputers, which feature many interconnected processors grouped together in so-called nodes, have become increasingly popular and powerful in recent years for AI research. The US Department of Energy’s top, which is the fastest supercomputer in the US and the second fastest in the world, has been used to help investigate things like unknown proteins. A few large tech companies, such as Microsoft and Nvidia, also have supercomputers for their own use.
In a Monday blog post, co-authored by Sengupta, Meta said that as of January its supercomputer includes 6,080 GPUs across 760 nodes, potentially making it one of the most powerful in the world. At its current size, Sengupta said, it’s on par with the National Energy Research Scientific Computing Center’s Perlmutter supercomputer, which ranks fifth in the world. When completed later this year, Meta said it will have 16,000 GPUs, and the company expects it to be capable of nearly five exaflops of computing performance, or 5 quintillion operations per second. .
“When fully built, the cluster will nearly triple in size, which we know will make it the fastest AI supercomputer in the world,” Sengupta said.
Meta said its researchers have begun using the supercomputer to train large models related to natural language processing and computer vision, and researchers will be able to use the supercomputer to “seamlessly analyze text, images, and videos together” and offer new augmented reality tools. Over time, Meta hopes this will allow the company to “build entirely new AI systems” that can perform hard computational tasks such as real-time translations for a large group of people who all talk about different languages.
The company said early tests showed the supercomputer could train large language models three times faster than the system it currently uses. That means an AI model that would take nine weeks to train with the existing system could be trained in three weeks with the supercomputer.
Meta hopes its new supercomputer will eventually be able to train AI models with trillions of parameters – there are only a few known existing AI models at this scale. It would also be several times larger than GPT-3, which is a large language model from OpenAI that can generate human-sounding text and is used for applications such as language learning and tax software for freelancers. .
Eventually, Meta said, the supercomputer will lead to the technologies needed to build a so-called “metaverse” – a vast, interconnected virtual realm that people can wander through, via digital avatars, and interact with others who are also there virtually. Facebook positioned the metaverse as the future of the company, but today only snippets of that vision have been realized.
Meta builds an AI supercomputer
| Local Business News Yahoo news
cnn World Gt
Not all news on the site expresses the point of view of the site, but we transmit this news automatically and translate it through programmatic technology on the site and not from a human editor.