Microsoft dedicated hundreds of millions of dollars toward building a supercomputer to power OpenAI projects such as ChatGPT, according to Bloomberg.
The supercomputer linked tens of thousands of Nvidia A100 graphics cards on Microsoft Azure to enable OpenAI to train its AI models.
- The system architecture was reliable and operational "at a very large scale," resulting in ChatGPT and other AI models, said Nidhi Chappell, Microsoft general manager of Azure AI infrastructure.
- While ChatGPT may be the most well-known model to come from the system, the supercomputer was built to be generalized for anyone to train large language models, according to Microsoft EVP Scott Guthrie.
- Overall, the project cost several hundred million dollars. So far, Microsoft has dedicated over $13B in total investments into OpenAI, including a $10B investment announced in January.
- Since 2019, Microsoft has been the exclusive cloud partner for OpenAI, and its Azure-powered supercomputing system has been used to train all of OpenAI's machine-learning and AI models.
- According to Guthrie, Microsoft is now working on a next-generation supercomputer that "is much bigger and will enable even more sophistication."
- Meanwhile, Microsoft is hosting an event, "The Future of Work: Reinventing Productivity with AI," this Thursday, March 16.
- The hosts, CEO Satya Nadella and Microsoft 365 lead Jared Spataro, are expected to show viewers "how AI will power a whole new way of working for every person and organization."
- The live stream will begin at 8 a.m. PT/11 a.m. ET.
- Relatedly, Microsoft recently announced a "Copilot" version of its Dynamics 365 business management platform, which aims to automate some customer service, marketing, and sales tasks.
- Dynamics 365 Copilot, described as an AI assistant, incorporates OpenAI technology to automate tasks, such as writing email responses, drafting chat and email answers, and producing email summaries or Teams meetings in Outlook.ش
