Meta is building a gigantic supercomputer called AI Research SuperCluster to train artificial intelligence models. Meta claims that it would be the fastest Super Computer in the world when it’s completely prepared in mid-2022.
Meta said it will have 16,000 GPU’s (graphical processors) – and will be capable of five exaflops of computing performance — that’s 5 quintillion operations per second.
Meta said its researchers have begun using the supercomputer to train large models related to natural-language processing & computer vision – and researchers will be able to use the supercomputer to “seamlessly analyze text, images & video together” and come up with new augmented reality tools.
In this post, we will present a brief description of a massive supercomputer named AI Research SuperCluster.
The AI Research SuperCluster is 20x faster than Meta’s V100-based production clusters. Similarly, it is 9x faster than Meta’s V100-based research clusters. It is well-found with trillions of parameters for natural language processing (NLP) and training models. It may effortlessly study images, texts, and videos through hundreds of different languages.
According to the Meta, its researchers have now started using RSC to train large models in natural language processing (NLP). They have trained big models of computer vision for research. Their main aim is one-day training models with trillions of parameters.
The company thinks that RSC will help its AI researchers build new and improved AI models. Those may learn from trillions of instances. They can also work across hundreds of different languages, effortlessly examine text, images, and video together. Similarly, they can develop new augmented reality tools and much more.
Advanced AI and Machine Learning Computations
The researchers say that they will be able to train the largest models required to develop forward-thinking AI for computer vision, NLP, speech recognition, and more.
The Research SuperCluster will support us to build completely new AI systems. For example:
- Power real-time voice translations to large groups of people
- Each speaks a different language.
- Therefore, they may flawlessly work together on a research project.
- Play an AR game together.
- Eventually, the work completed with RSC will head toward building technologies for the next main computing platform 0 the metaverse.
- Where AI-driven applications and products will play a key role.
- We need to understand that together supercomputers and AI supercomputers make calculations using what is identified as floating-point arithmetic.
- That is a mathematical shorthand.
- It is very valuable for making calculations using very big and very small numbers.
- The floating-point in question is the decimal point that floats between important figures.
- The degree of exactness deployed in floating-point calculations may be used based on different formats.
- The speed of the greatest supercomputers is designed using what are known as 64-bit floating-point operations per second.
- On the other hand, as AI calculations need less accuracy, AI supercomputers are frequently measured in 32-bit or even 16-bit FLOPs.
- That’s why relating the two types of systems is not certainly apples to apples.
- However, this caution doesn’t diminish the incredible power and size of AI supercomputers.
For more details visit:https://www.technologiesinindustry4.com/2022/01/ai-research-supercluster.html