Facebook group wants to build the most powerful supercomputer

Status: 01/25/2022 3:19 p.m

The Facebook group Meta announces the most powerful supercomputer in the world for artificial intelligence. It should be finished this summer. Does the calculator also help against hate online?

“AI Research SuperCluster” is the name of the supercomputer on which researchers from the Facebook group Meta have been working since 2020. It should be finished this summer. This was announced by the Silicon Valley company. Then the supercomputer should take over applications with artificial intelligence for the Internet company.

For example, it is planned that the computer will enable real-time translations that larger groups of people can use to converse in different languages. Another task of the supercomputer should be the detection of hate messages and posts on Facebook and Instagram, the group said.

Technology for the “Metaverse”

So far, Meta has already used artificial intelligence to automatically filter out content that is prohibited on its platforms. According to the company, 98 percent of hate speech is discovered before anyone sees it.

Meta devotes special energy to the development of new augmented reality tools and technologies for the so-called “Metaverse”. In this virtual world, in which AI-based applications are likely to play a special role, today’s social online experiences should merge with the real world, according to the company. This is where the group sees the future.

Also cooling developed by Facebook

The AI ​​Research SuperCluster (RSC) was developed from scratch by the Facebook group, including the solutions for cooling and cabling. As early as 2013, Facebook founded the research center in which scientists research, among other things, how to advance the learning process of AI models. According to the company, the first generation of such a computer was already developed in 2017.

The RSC is currently in operation with currently 6080 graphics processors. Fully expanded by mid-2022, it will connect 16,000 graphics processors. They are considered to be particularly efficient for computing when learning artificial intelligence. Meta uses chips from market leader Nvidia for its supercomputer.


Leave a Comment