Meta, the parent company of Facebook, has announced it is in the process of building an “AI supercomputer” expected to be done by mid-2022.
What this means is the company has created a computer, coined AI Research SuperCluster (RSC), with the purpose of training machine learning systems that is expected to be the fastest in the world.
What was once a far-off concept, AI has taken the front seat in companies’ operational strategies. Firms such as Microsoft have also created their own AI supercomputers.
However, this technology differs from regular supercomputers. The RSC will have the ability to train systems across Meta’s businesses, such as being able to detect hate speech on Facebook and Instagram, as well as design the company’s metaverse.
The latter is a recurring theme in the company’s future plans, in which it is hoping to create its own advanced digital universe where people can work, interact, make purchases, and more. The metaverse has already seen increased competition from rivals like Microsoft and Google since the beginning of the year.
Phase one of RSC is currently operating, which Meta says is improving its performance on its standard machine vision research tasks by up to 20 times.
Phase two is expected to be complete by the end of 2022, where the company hopes it will be able to train AI systems “with more than a trillion parameters on data sets as large as an exabyte.”
“RSC will help Meta’s AI researchers build new and better 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,” wrote Kevin Lee and Shubho Sengupta, engineers at Meta. “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.”
