US, WASHINGTON (ORDO NEWS) — Microsoft has developed a new Azure cloud-based supercomputer, a non-profit research organization OpenAI, founded by Elon Musk and Sam Altman. The company introduced a supercomputer at the largest technology conference Microsoft Build 2020.
The supercomputer is in the top five list of the most powerful supercomputers in the world based on open data from the TOP 500 Supercomputers rating. It was created to train a new generation of neural networks that require huge computing resources.
The difference between the new artificial intelligence models is that the tasks will be performed by one large model, trained on a huge amount of data. Previously, AI models were relatively small and could use sets of labeled data to perform any one separate task – to recognize an object, translate text, etc.
Microsoft says that a large model can, for example, better understand the nuances of speech, grammar, concepts and context, and is also able to perform several tasks: summarize long speeches, moderate chat in online games and even generate program code based on search results on github.
The creation and training of such massive models requires a progressive supercomputing infrastructure or clusters of modern equipment connected through high-bandwidth networks. Since Microsoft’s supercomputer is powered by the Azure cloud, it has the benefits of a modern cloud infrastructure, including rapid deployment, data center reliability, and access to Azure services.
The supercomputer has 285 thousand processor cores (CPU core), 10 thousand graphic processors (GPUs) and network bandwidth of 400 gigabits per second for each GPU server.
OpenAI – an organization that proclaimed its goal to create humanity-friendly artificial intelligence – has become famous as a developer of bots for the tactical strategy of Dota 2, able to beat humans. In order to learn what was unthinkable a few years ago, AI systems spent 45 thousand years in a “DotA” in total, playing against each other in accelerated mode. In addition, OpenAI created the fake text generator GPT2, but refused to publish the full results of the study, believing that the AI model could be harmful, because the false news it produces seems too convincing.
OpenAI has the ultimate goal of creating an artificial general intelligence (AGI) system that will be able to “think” just like a person.
According to most scientists, AGI systems that are able to learn, reflect and excel in solving intellectual problems will appear no earlier than in decades or even centuries, if this idea is achievable at all. Despite this, the world’s richest companies are investing in the development of a “strong” AI. The same goal, for example, pursues the laboratory DeepMind, owned by the owner of Google.
Some representatives of the IT industry, including Elon Musk and Peter Thiel, have invested less than $ 1 billion in OpenAI to study artificial intelligence. True, Musk resigned from the board of directors of OpenAI, explaining this by a divergence in views on the development of the organization. Until 2020, the company planned to spend only a small part of the first investment, however, it seems that a little more was spent.
One of the founders of OpenAI told Bloomberg that the company needed a partnership with Microsoft, including due to a lack of computing power. According to him, from 2012 to 2018, the number of calculations needed to run artificial intelligence models has grown more than 300 thousand times.
Contact us: [email protected]