Home  >  Article  >  Technology peripherals  >  GPT-4 has been working together for many years and is still being trained on Microsoft AI supercomputers

GPT-4 has been working together for many years and is still being trained on Microsoft AI supercomputers

WBOY
WBOYforward
2023-04-12 15:25:241218browse

According to foreign media reports, ChatGPT has gained a large number of users since its launch on November 30 last year. It exceeded the 100 million mark in January. With the increase of attention and the increase of supported devices, the number of users continues to increase. .

已合作多年 GPT-4仍在微软AI超级计算机上训练

OpenAI, which has been very successful on ChatGPT, has not stopped their work on large language models Continuing their exploration, they launched a new generation of natural language processing model GPT-4 on Tuesday local time.

According to information announced by OpenAI on its official website, the newly launched GPT-4 is their most advanced system at present, with more accurate problem-solving capabilities and superior reasoning capabilities compared to those launched in November last year. ChatGPT has over 100 million users.

Like the previously launched GPT, GPT-2 and GPT-3 large language models, OpenAI also relies on the infrastructure of its partner and important investor Microsoft for the newly launched GPT-4.

OpenAI stated on its official website that GPT-4 was trained on the artificial intelligence supercomputer of Azure, Microsoft’s cloud computing business unit. Microsoft Azure’s artificial intelligence optimized infrastructure also allows them to provide it to users around the world. GPT-4.

It is worth noting that the day before OpenAI launched GPT-4, Microsoft had disclosed on its official website that they and OpenAI began to cooperate in 2019. After reaching a cooperation agreement, the two companies began to work on Azure We work closely with the company to build supercomputing resources that are designed and dedicated to allow OpenAI to train an increasingly powerful suite of AI models.

The supercomputing facility they built uses tens of thousands of NVIDIA GPUs optimized for artificial intelligence, and are connected together through a high-throughput, low-latency network based on NVIDIA quantum InfiniBand communication. High performance computing.

The above is the detailed content of GPT-4 has been working together for many years and is still being trained on Microsoft AI supercomputers. For more information, please follow other related articles on the PHP Chinese website!

Statement:
This article is reproduced at:51cto.com. If there is any infringement, please contact admin@php.cn delete