Research teams and Microsoft Azure are building an all-new AI infrastructure service called the ‘Singularity.’
Engineers and coders at Microsoft and Microsoft Azure are working to create a new AI infrastructure service by the name of ‘Singularity.’ This service is about providing AI practitioners and data scientists with a way to scale, build, iterate, and experiment on their models based on Microsoft provided and distributed infrastructure specially designed or the AI.
A group of people working on the project has also published a paper on Singularity, wherein they say that:
“At the heart of Singularity is a novel, workload-aware scheduler that can transparently preempt and elastically scale deep learning workloads to drive high utilization without impacting their correctness or performance, across a global fleet of accelerators (e.g., GPUs, FPGAs).”
Many of the officials at Microsoft have said that their plan of action has always been to make the training optimization tools and large AI models of the company available via GitHub and Azure AI services. Microsoft also creates multiple services and accelerators under the banner of ‘Azure AI’ for customers that do not require a devoted supercomputer. This news comes only two months after Microsoft announced its AI expansion into supercomputers with the Azure 80GB NVIDIA A100 GPU.
Codename: ‘Singularity’ – the go-to name
This is not the first time Microsoft has used the codename ‘Singularity’ for a research project. Nevertheless, their previous Singularity experiment was basically a microkernel operating system, with related libraries and toolsets formed entirely in managed code. That Singularity was not Windows-based, either, and was created from scratch as a concept proof. But it did influence or spawn various other Microsoft OS research projects, including Helios, Barrelfish, Drawbridge, and Midori.
After all their efforts, though, Microsoft is not the only tech organization working to make AI supercomputing abilities available to customers. Meta is also wants something similar, and it came as no surprise when it launched the metaverse as a building block of their supercomputing vision.
9 hours ago
