Microsoft works with OpenAI to exclusively license GPT-3 language model
At the Ignite 2020 conference held this week, Microsoft announced that it is cooperating with OpenAI and has obtained an exclusive license for the GPT-3 language model.
GPT-3 is by far the largest and most advanced AI language model in the world, consisting of 175 billion parameters. Microsoft said the model has been trained on Azure’s AI supercomputer. Next, OpenAI will continue to provide GPT-3 and other AI models through its own API hosted on Azure.
In July last year, Microsoft invested $1 billion in OpenAI to reach a partnership. In exchange, OpenAI agreed to grant part of its intellectual property rights to Microsoft and then packaged and sold by Microsoft. The two parties also cooperated in the development of new Azure AI supercomputing technology, and Microsoft has become the only cloud provider of OpenAI.
At the Build 2020 developer conference in May this year, the AI supercomputer hosted by Azure made its debut. It contains more than 285,000 processor cores, 10,000 graphics cards, and has 400 gigabits per second connection per graphics card server ability.
Microsoft’s executive vice president and chief technology officer Kevin Scott repeatedly emphasized the development trend and importance of AI in blog posts. With the exclusive authorization of GPT-3 this time, Scott believes that Microsoft can use this technology to develop and provide more advanced AI solutions. In his view, the commercial and creative potential released through GPT-3 is immeasurable, and its achievable uses are even far beyond people’s imagination.