Meta has launched a supercomputer to combat hate speech on Facebook and Instagram
Mark Zuckerberg’s company Meta has announced the launch of the world’s fastest artificial intelligence supercomputer. The main task of this supercomputer is to provide artificial intelligence systems in the metacomplex that Zuckerberg is creating. In particular, the computer now monitors the use of hate speech on social networks such as Facebook and Instagram.
Zuckerberg’s predictions
The company’s brainchild is called RSC, or AI Research SuperCluster. Mark Zuckerberg promises that the main phase of work on the supercomputer will be completed by the end of 2022. Allegedly, by then the machine will be the fastest in its class. The main use of the supercomputer is to create and coordinate artificial intelligence systems in VR and AR.
Meta’s supercomputer capabilities
Meta’s brainchild is by no means the first supercomputer created in the world. Almost all of the world’s leading universities and research organisations have similar machines. But the fundamental difference between a supercomputer and Meta is that the specifics of its operation do not require it to solve problems with the degree and meticulousness required in science. This allows Meta’s supercomputer to achieve astonishing performance.
Meta started working on the RSC a year and a half ago. To date, the computational complex consisting of 760 Nvidia GGX A100 systems containing 6080 GPUs has been launched. It is expected that by the end of 2022, the number of GPUs combined will reach 16,000. The result will be a huge computing complex, training artificial intelligence systems on more than a trillion parameters. The total volume of data to be processed will reach an egabyte.
RSC will mainly be used to train other systems in Meta, to provide functions of augmented reality. One of the most important spheres of supercomputer’s usage will be revealing the facts of using hate speech on Facebook and Instagram, and also suppressing the attempts to use these social networks for spreading deliberately untrue information, which can destabilize public processes in one or several countries.
The bugs are still there.
However, experts acknowledge that, to date, the content moderation algorithms on the aforementioned social networks have been extremely inefficient and seriously flawed, resulting in high-profile scandals such as the Facebook Papers, among others.
Nevertheless, the company is not giving up and is determined to create fundamentally new artificial intelligence systems that will work with content created in almost all modern languages with writing, analyse multimedia content, and support augmented reality tools.
In particular, Meta promises that the supercomputer will become an assistant in solving such problems as: providing real-time voice translation, providing users with tools to work on research projects, as well as supporting super-advanced AR games.