Facebook today announced a new server design it calls Big Basin, a successor to its Big Sur line of artificial intelligence training systems. These Nvidia-powered GPU servers, tied together into large training networks for AI software, are what enable Facebook products to perform object and facial recognition and real-time text translation, as well as describe and understand the contents of photos and videos. Big Basin can now train on learning models 30 percent larger than its predecessor, Facebook says. It can also crunch through the massive number sets used by an AI system to improve itself at nearly twice the speed, according to tests conducted on standardized neural network models. Facebook plans to make the server design open to the public in the near future. That’s standard at the company, which participates in and helped create the Open Compute Project for sharing and collaborating on data center hardware and software. So anyone — even server design specialists in competing companies — will soon be able to download the Big Basin schematics once they’re posted online.
For Facebook, it’s less about keeping under wraps the tools it uses to train AI systems and more about trying to advance what its AI systems are capable of. It’s not just about pushing the limits of technology, though Facebook is among one of the largest organizations investing in cutting-edge and experimental AI research. The company’s large investments in AI go hand-in-hand with its push toward live video and other consumer-centric focuses. “If you’ve logged into Facebook, it’s very likely you’ve used some type of AI system we’ve been developing” says Kevin Lee, a technical program manager at Facebook who works on Big Basin and other data center initiatives. For instance, by tagging friends and categorizing videos — including those streamed live — Facebook may help drive more users to upload video and consume it. There’s also a large social impact the company can have with its AI research. One key function of Facebook’s current AI algorithms today is describing the contents of photos to blind users, while just last week Facebook announced it would use AI-powered pattern recognition software to try and identify when troubled users may be in need of mental health outreach. All of this is made possible because the company invests in and continues to develop the servers, like Big Basin, that train these systems before they’re pushed out to public products.
With the kind of dedication and finances invested by Facebook in making technology open source, that time is not far when it’ll be one of the most loved company amongst the development community.