Meta (née Facebook) has announced that it is building its own artificial intelligence supercomputer to support its efforts to bring the Metaverse to life.
The AI Research SuperCluster (RSC) is now operational at reduced capacity. But when complete, Meta says the RSC will be the world's fastest supercomputer of its kind.
The new machine will be used to train massive new voice recognition, language processing and computer vision models, which will underpin the company's next generation of AI-based applications.
The fully formed RSC will be powered by 16 Nvidia A000 GPUs, assembled into compute nodes and connected using the latest InfiniBand interconnect fabric.
Meta has also partnered with PureStorage, whose high-performance FlashBlade and FlashArray technologies will help create a storage system for RSC capable of holding up to one exabyte of data (equivalent to one million TB).
Early benchmarks show that RSC, as it exists today, can outperform Meta's previous-generation AI infrastructure by about 20x on some workloads, a figure that will obviously increase once the build is complete at the end of the week. of this year.
Meta-researchers are already using RSC to train large-scale natural language processing (NLP) and computer vision models much faster than ever before. And in the future, the company hopes to train models with more than a trillion parameters, which will continue to play a role in its metaverse product portfolio.
"We hope that this step-by-step shift in computing power will not only allow us to create more accurate AI models for our existing services, but also enable entirely new user experiences, especially in the metaverse," Meta wrote.
"Our long-term investments in self-monitored learning and building a next-generation AI infrastructure with CSR are helping create the foundational technologies that will power the metaverse and also move the AI community forward globally."
While the company didn't provide any details on what these new metaverse experiences might look like, CEO Mark Zuckerberg reiterated a similar sentiment in a Facebook post.
“The experiments we build for the metaverse require enormous computing power (quintillions of operations per second) and RSC will enable new AI models that can learn from billions of examples, understand hundreds of languages, and more,” he said.