In a series of blog posts, Meta Platforms Inc., the parent company of Facebook, has recently shed more light on its ambitious AI (Artificial Intelligence) programs, revealing its efforts to harness AI technologies in new, exciting ways.
Meta’s Progress in AI Hardware
Meta’s commitment to accelerating its AI capabilities is evident in its creation of the MTIA (Meta Training and Inference Accelerator) – a silicon chip designed specifically for AI models, initially conceived in 2020. This chip, a complement to GPUs, was utilized by Meta to enhance recommendations, with subsequent iterations being developed to meet higher specifications.
Moreover, Meta has unveiled its new data center blueprint optimized for AI, intended to power AI applications, including inference and training processes. This optimized data center complements another AI-centric innovation, the Meta Scalable Video Processor (MSVP) – a chip specifically designed for video transcoding.
The Giant Leap: Meta’s AI Supercomputer
In a considerable leap towards AI research enhancement, Meta disclosed its plans of constructing an AI supercomputer embedded with 16,000 GPUs. The company proudly touted this technological marvel as being among the world’s most powerful AI supercomputers.
Notably, Meta’s ambitious AI advancements have been largely kept under wraps. The company had previously announced AI products intended for public use, but detailed information about these AI technologies remained undisclosed.
Meanwhile, the supercomputer, coined as Research SuperCluster (RSC), established in cooperation with Penguin Computing, Nvidia, and Pure Storage, has concluded its second development phase. This monumental effort signifies Meta’s resolution to expand its footprint in the artificial intelligence landscape. Focusing on developing custom AI chips and powerful supercomputers, Meta demonstrates its commitment to be meat the forefront of AI advancements.
Throughout the current year, Meta has been vocal about AI-powered advertising initiatives and has highlighted the usage of AI in aspects such as content discovery and user support.
The company’s AI language model, LLaMA, made unexpected headlines in March when it was leaked, granting the open-source community unintended access to the technology. This unintentional leak underscored Meta’s heavy involvement in cutting-edge AI research and development, affirming the company’s drive to take AI capabilities to the next level.