Meta: New supercomputer enables revolutionary AI

Meta wants to expand Facebook’s capabilities with the help of a powerful supercomputer. The so-called “AI Research SuperCluster” is said to hold immense potential, especially in the field of artificial intelligence.

Nvidia technology as a basis

As an average consumer, we can hardly imagine a computing power of 5 exaflops. But the so-called “AI Research SuperCluster” is not a conventional computer either. It deserves its name as a supercomputer not only because of its sheer computing power. Its capabilities in the field of artificial intelligence are supposed to be particularly exciting and promising. The 5 exaflops are no coincidence. A total of 760 Nvidia DGX-A100 systems are to be able to provide the corresponding power in cooperation with 16,000 A100 accelerators before the end of this year.

Training for AI models

But what has Meta created this huge data center for? Once again, the answer is artificial intelligence. Obviously, this is the key to success for Meta and Facebook when it comes to the future. However, it requires immense computing power to be able to train the corresponding models. This should definitely be available with the supercomputer. Meta demonstrates this to its users with some impressive figures. It should be possible to perform special model calculations with a maximum of 1 trillion different parameters. On top of that, 1 exabyte large data bundles are no problem for the “AI Research SuperCluster”. What may seem frightening at first glance should perhaps ensure that postings on the social network can be better monitored from now on. In particular, Meta might then be able to better combat dangerous postings with content relevant to criminal law in the future.