Microsoft has developed a new supercomputer on the basis of the Azure cloud non-profit research organization OpenAI founded by the Elon Musk and Sam Altman. Supercomputer, the company introduced the largest technology conference Microsoft Build 2020.
the Supercomputer is in the top five of the list of most powerful supercomputers in the world based on open data of the TOP 500 Supercomputers. It was established for training a new generation of neural networks, which require huge computational resources.
unlike the new models of artificial intelligence is that the tasks will run one big model, trained on enormous amounts of data. Earlier AI models were relatively small and could use a set of annotated data to perform any one individual task – to identify the object, translate text, etc.
Microsoft says that a large model can, for example, more subtly understand the nuances of speech, grammar, concepts and context and is able to perform several tasks: to summarize long speeches, to moderate the chat in online games and even generate the code for the search results on GitHub.
the Creation and exercise of such massive models requires progressive superficialities infrastructure or clusters of modern equipment, coupled with networks of high bandwidth. As a supercomputer developed on the basis of Microsoft Azure cloud, it has the advantages of modern cloud infrastructure, including rapid deployment, reliability of data centers and access Azure services.
In the supercomputer 285 thousand processor cores (CPU core), 10 of thousands of graphics processing units (GPUs) and network bandwidth of 400 gigabits per second for each GPU server.
OpenAI — an organization that proclaimed itself the task of creating friendly to mankind of artificial intelligence, — Prohe was known as a developer of bots for tactical strategy Dota 2, who can win the person. To learn things that were unthinkable just a few years ago, AI systems total spent in “DotA” 45 thousand years playing against each other at an accelerated rate. In addition, OpenAI have created a generator create fake texts GPT2, but refused to publish full results of the study, considering that the AI model may cause harm, because it issued false news seem too convincing.
OpenAI HSUS ultimate goal is to create a system of artificial General intelligence (AGI), which will be able to “think”, as well as people.
According to most scientists, the AGI system that is able to learn, think and outperform humans in the solution of intellectual tasks, will appear not earlier than in decades, or even centuries, if at all achievable. Despite this, the development of “strong” AI invested richest companies in the world. The same goal, for example, pursues laboratory DeepMind owned by the owner of Google.
Some representatives of the IT industry, including Elon Musk and Peter Thiel, invested in OpenAI less than $1 billion to study artificial intelligence. However, Musk left the Board of Directors of OpenAI, citing a divergence of views on the development of the organization. By 2020, the company planned to spend only a small part of the first investments, however, seem to spend a little more.
One of the founders of the OpenAI told the publication Bloomberg that the company needed a partnership with Microsoft, including due to lack of computing power. According to him, from 2012 to 2018, the number of calculations required to run the models of artificial intelligence has grown more than 300 thousand times.