Meta is the last of the big tech to have built its own supercomputer AI and, according to what says Zuckerberg in a statement, by the end of 2022 it will be the fastest fastest in the world. Meta's rivals, Microsoft and Nvidia, have already announced theirs and the fight for supremacy in the field of artificial intelligence seems tighter than ever.
This type of supercomputer, also called RSC (AI Research SuperCluster, supercluster of artificial intelligence research), will be used to train the artificial intelligence in different fields, so you can use it to moderate content on Facebook and Instagram , for translations in real time, simulations with viewers for augmented reality and above all for the creation of the much anticipated Metaverso , the company's brand for interconnected virtual spaces, from offices to the stadiums.
Meta's new RSC , when completed, will consist of 16,000 GPUs (a type of processor particularly suitable for dealing with machine learning ), much more than Microsoft's, which contains 10,000. These AI supercomputers are similar to datacenters, like normal supercomputers, in the sense that they are composed of a very high number of interconnected processors , but they differ from the latter in that artificial intelligence has a greater need for speed in calculations , rather than precision. So their speed instead of being measured in FLOPs ( floating point operations per second , floating point operations per second) at 64 bit, it is measured at 32 or even 16 bit.
So the speed of supercomputers and RSC cannot be compared , because their usefulness must be considered for what they are designed to do. In this sense, the Meta project takes place at a time when trust in society is at historical lows and is also a way to rebuild the relationship with its users.
