Meta Corporation has reportedly finally managed to complete the first phase of development of a proprietary supercomputer with its own artificial intelligence. And going into more detail, soon after the AI Research SuperCluster is fully built at the end of this year, the company believes that it will become literally the fastest AI supercomputer on the planet, which is capable of “performing almost 5 exaflops of computing with mixed accuracy. Moreover, in this regard, right now the company says that the AI Research SuperCluster will even be able to help researchers develop better artificial intelligence models that can learn from trillions of different examples.
It should also be noted that, according to the Meta Corporation, among other things, the artificial intelligence models present in the above-mentioned supercomputer will be able to create much more advanced augmented reality tools, as well as “seamlessly analyze text, images and videos together.” Much of this work serves to help achieve and realize a vision for a metaverse in which AI-based applications and products will play a key role, according to the corporation.
“We hope that the AI Research SuperCluster will help us create completely new artificial intelligence systems, which in turn can, for example, provide real-time voice translation for large groups of people who speak different languages at the same time so that they can collaborate seamlessly, for example, in various research projects, and to enjoy all the features and benefits of augmented reality together with other people,” wrote Kevin Lee and Shubho Sengupta, technical program manager and software engineer at Meta, respectively, in their blog. Be that as it may, the actual AI Research SuperCluster setup currently has 760 Nvidia DGX A100 systems with 6080 GPUs from the aforementioned company. And yes, Meta also believes that the current version of the project is already among the fastest artificial intelligence supercomputers that exist on the planet.
And based on early tests, the company claims that the AI Research SuperCluster can, compared to the company’s old installation, run computer vision workflows up to 20 times faster, while the NVIDIA Collective Communications Library will run more than nine times faster. faster. Also, do not ignore the message that the AI Research SuperCluster will also be able to train large-scale natural language processing models three times faster. Thus, AI models that in turn determine whether “an action, sound or image is harmful or harmless”, for example, in order to eliminate hate speech, can be trained much faster.
