In an announcement that could well act as Microsoft’s defining moment in the future that we envision as AI, the Redmond giant announced today, a supercomputer that is among the top 5 globally. Mulled by Microsoft among “one of the top five publicly disclosed supercomputers in the world”, it has been designed in partnership with and exclusively for OpenAI, in what is arguably company’s best attempt yet, at developing advanced AI processing for a computer. Microsoft and OpenAI have partnered to give birth to Azure AI computing technologies wherein Microsoft has contributed a whopping $1 billion.
The supercomputer has been exclusively designed to train OpenAI’s artificial intelligence models. Microsoft also aims to provide the basic infrastructure that would be needed by other developers giving access to a brand new platform. Kevin Scott, Chief Technical Officer, Microsoft emphasised the importance of variety that these models are bringing to the table. “The potential benefits extend far beyond…one type of AI model” he added.
Microsoft has also brought to life, the latest version of DeepSeed (an open source deep learning library for PyTorch) which claims to reduce the amount of large distributed model training’s computing power.
Without going into a lot of depth about the processing it was only detailed how the supercomputer has an access to Azure services and is “a single system with more than 285,000 CPU scores, 10,000 GPUs and 400 gigabits per second of network connectivity for each GPU server.” OpenAI has gone to the maximal length of describing Microsoft’s design as their “dream system”. Compared with other machines listed on the TOP500 supercomputers in the world, it ranks in the top five, Microsoft says.
Scott further mentioned how unimaginable newer applications are possible with these latest sets of combinations with a multitasking class of AI models which might just outsmart humans. There AI research community has been successful in discovering an integration of smaller AI models, an alloy of essential task performers, resulting in the formation of a singular giant of a model that is so deeply absorbing in content that its advantages are unparalleled and countless, all to achieve optimisation and paramount efficiency.
“This is about being able to do a hundred exciting things in natural processing at once and a hundred exciting things in computer vision” he added enthusiastically which only makes us anticipate the launch of this supercomputer, relying on his words.
For AI fanatics without the requirement of a full fledged supercomputer, Azure AI has also come up with complete access to “powerful compute” with all the AI accelerators as they are present in the supercomputer. Microsoft also brings us the “Turing mode” which uses “self-supervised” learning. In the course of its training this model, the aim will be to cover the gap between missing words and even form a synopsis of longer speeches. Earlier this year, it also released to researchers the largest publicly available AI language model in the world, the Microsoft Turing model for natural language generation.
The goal, Microsoft says, is to make its large AI models, training optimization tools and supercomputing resources available through Azure AI services and GitHub so developers, data scientists and business customers can easily leverage the power of AI at Scale.