Facebook’s parent Meta says it has created a powerful supercomputer that it expects to be the world’s fastest when it is completed later this year.
The company recently announced the computer will be used to create better artificial intelligence (AI) models and to improve operations that process huge amounts of data.
Supercomputers are made up of hundreds or thousands of powerful machines. They work together to perform complex operations that are not possible with normal computing systems.
Meta’s supercomputer is currently operating. The company calls it the AI Research SuperCluster, or RSC. In a statement, the company said RSC “will help Meta’s AI researchers build better AI models that can learn from trillions of examples.”
The company did not provide information on where the computer is kept or how much money is being spent to build it.
An AI system can be trained to recognize different kinds of content and perform actions examining huge amounts of data. Such systems require very powerful computers.
Meta said its supercomputer will be able to process written information, images, and video that is in hundreds of different languages. Meta said the computing system can process images and video up to 20 times faster than current systems. The company is expected to use the system for its Facebook, Instagram, and WhatsApp services.
The company said it believes RSC is currently one of the world’s five fastest supercomputers. It said RSC would become the fastest when it is fully operating.
In addition, Meta said it hopes RSC will help in the development of completely new AI systems. One example is a planned tool that would permit real-time voice translations to large groups of people interacting in business or social situations.
Such an operation fits with Meta’s stated goal of developing technologies to be used in a future “metaverse.” Metaverse is a term used to describe a non-physical world in which individuals interact through different kinds of digital technology.
Meta chief Mark Zuckerberg wrote in a Facebook message: “The experiences we’re building for the metaverse require enormous compute power (quintillions of operations [a] second!)…”
Meta uses a combination of human moderators and AI to identify and remove, what it considers, harmful content and misinformation. Some critics say the company is not doing enough to block bad content. It has also faced government scrutiny over its privacy and business methods in the United States and Europe.
In a statement, the company said its supercomputer will be helpful in “critical use cases like identifying harmful content.”
Diego Naranjo is the head of policy for European Digital Rights. It is a group of non-governmental organizations seeking to restrict the power of large technology companies. He told the French press agency AFP that he recognizes Facebook has made some efforts to protect users’ privacy and limit harmful content.
But he questioned what the company might do with such a powerful new tool. “Nothing good can come from all that computer power in the hands of such a tech superpower,” Naranjo said.
Meta said its supercomputer will use “real-world examples” from its own systems during the process to train its AI.
“They are going to, for the first time, but their customer data on their AI research computer,” said Tuomas Sandholm. He is a computer science professor and co-director of the AI center at Carnegie Mellon University in Pittsburgh. Sandholm said it marked “a really big change” for Facebook to give its AI researchers and computing systems access to all that data.
Meta said the data used to train AI models will go through a “privacy review process” to ensure that it is not linked to individual users. It added that the data will also be encrypted before entering the training process.
