Artificial intelligence requires the simultaneous execution of a large number of low-precision calculations. And the GPUs are designed to have thousands of processing cores that can all work at the same time. Meta’s AI supercomputer currently houses 6,080 Nvidia graphics-processing units and will reach nearly as large as 16,000 GPUs by mid-summer this year, making it the fastest AI supercomputer in the world. The Facebook team of researchers have been developing this supercomputer for over two years and has now been utilizing the supercomputer to train AI models in natural-language processing and computer vision for research purposes
The experiences we’re building for the metaverse require enormous compute power…and RSC will enable new AI models that can learn from trillions of examples, understand hundreds of languages, and more.
Mark Zuckerberg, CEO of Meta
Several hundred employees were involved in the study, according to the business, including researchers from partners Nvidia Inc., Penguin Computing Inc., and Pure Storage Inc. With its aim to ingest troves of data to build AI models that can think like a human brain, with multiple inputs—such as voice and visual recognition—and can deliver a contextual understanding of situations.
The team aims to achieve more than a trillion parameters that can be trained on data sets as large as an exabyte, which is nearly 36,000 years of high-quality video. Apart from that, the key objective is to ingest troves of data to build AI models that can think like a human brain, with multiple inputs—such as voice and visual recognition—and can deliver a contextual understanding of situations.
For us, it’s an order of magnitude larger compute that’s available to our researchers to train a single model with a lot more compute than anybody else in the world has access to.
We believe that this will really unlock an AI that just understands the world around you much, much better.
Jerome Pesenti, vice president of AI at Meta
Regular supercomputers are optimized for high-precision activity, whereas AI supercomputers operate on much lower levels of precision, gaining speed without affecting the end results.
Shubho Sengupta, meta’s Lead software engineer of supercomputer ai project
High-performance computing infrastructure, such as a supercomputer, is necessary to make sense of the volumes of data that Facebook needs.
In Facebook’s case, I suspect what they’re looking at is the treasure trove that they have of the data that they get from their users, and that’s a huge data set, too big for most of the artificial intelligence systems that researchers typically use,
Bill Gropp, director of the National Center for Supercomputing Applications at the University of Illinois at Urbana-Champaign.
In the metaverse, it’s one hundred percent of the time, a 3-D multi-sensorial experience, and you need to create artificial-intelligence agents in that environment that are relevant to you.
Jerome Pesenti, vice president of AI at Meta
Read more about the news here.