A new supercomputer designed to accelerate AI research and development has been unveiled by Meta, a company specializing in Artificial Intelligence (AI) at the University of California, Berkeley.RSC will assist in the development of new and improved AI models, as well as work across hundreds of languages and develop new augmented reality capabilities.
The RSC is thought to be one of the world’s fastest AI supercomputers. The supercomputer will be capable of quintillions of operations per second. The RSC is being built as part of a multi-million-pound project to develop the next generation of AI technology. It will be among the fastest supercomputers in the world when fully built out in mid-2022.
RSC will assist Meta’s AI researchers in developing better AI models that can function across several languages and evaluate text, photos, and video simultaneously. The collaboration will pave the path for the development of technologies for the next major computing platform, the metaverse.
Since 2013, Meta has been working on AI development, including self-supervised learning, in which algorithms can learn from large numbers of unlabeled samples, and transformers, which allow AI models to reason by focusing on certain portions of their input. To fully exploit the benefits of advanced AI, numerous domains, such as vision, voice, and language, will necessitate the training of increasingly big and complex models, particularly for crucial use cases such as recognizing hazardous content. Meta concluded in early 2020 that the best approach to expedite progress was to create a new computing infrastructure—RSC.
Meta has announced the release of their RSC platform, which will aid in the training of models that employ multimodal signals to judge whether an action, sound, or image is dangerous or benign.