Meta has entered the supercomputer pissing contest and claims its new Research SuperCluster (RSC) will be the world’s fastest AI supercomputer when it is completed sometime around mid-2022. Wonderful.
That said, the “world’s fastest” isn’t necessarily as big a deal as Meta and Mark Zuckerberg would like you to believe. As The Verge explains, an “AI supercomputer” is a bit different than our traditional notion of a standard “supercomputer.” In a (very tiny) nutshell, the former is increasingly used in machine learning projects, and thereby doesn’t need to be as accurate as the latter technology. Not only that, but companies often like to tout their supercomputer’s top speeds as if that’s their consistent performance rate, when in reality that’s far from the case. In fact, they often barely clock in at 25 percent of their advertised speed when put to work on real projects.
Knowing all that should minimize the severity of our “Facebook Deathbot” nightmares later tonight...
Putting it to work — According to the company’s announcement, the House of Zuckerberg began construction on the project a year-and-a-half ago, and started completely from scratch. When completed later this year, Meta claims the RSC will be comprised of 16,000 GPUs for AI systems training purposes, “with more than a trillion parameters on data sets as large as an exabyte.”
“We hope RSC will help us build entirely new AI systems that can, for example, power real-time voice translations to large groups of people, each speaking a different language, so they can seamlessly collaborate on a research project or play an AR game together,” Meta engineers Kevin Lee and Shubho Sengupta wrote in a new blog post.
Bragging rights only go so far — That kind of advanced research is certainly cool to imagine, but it doesn’t do much to improve the company’s image as it continues to tank in the public’s eye. Despite their very (very) concerted push to make the metaverse a thing, Zuckerberg and crew still face mounting federal pressure on antitrust issues, along with ire from both the political spectrum’s left and right sides for the spread of misinformation, perceived bias, and privacy concerns. Even the world’s fastest AI supercomputer can’t outrun the company’s approaching consequences.