Microsoft will expand its partnership with OpenAI, an AI research organization founded in 2015 by Elon Musk to exclusively license the group's powerful GPT-3 language model. The move, revealed during Microsoft Ignite 2020 virtual conference, will enhance Microsoft's already powerful suite of natural language tools.
Last year, Microsoft revealed a $1 billion investment in OpenAI, months after the research organization became a for-profit company. With the investment, Microsoft aimed to commercialize OpenAI projects, develop new Azure AI supercomputing technologies, and further research into artificial general intelligence.
Better language models
By gaining exclusive licensing to OpenAI's GPT-3 language model, Microsoft will strengthen its existing language models, seen as some of the most advanced on the market.
"Microsoft has reached what it claims as human parity in major fields such as language and speech and this year we have seen it start to communicate on its massive, at-scale model," said Nick McQuire, senior vice president and head of AI and enterprise research at CCS Insight, referring to Turing-NLG, a deep learning language model with more than 17 billion parameters.
Turing-NLG was the largest language model available when Microsoft revealed it in February of this year. According to Microsoft at the time, Turing-NLG could enhance writing assistance capabilities within the Microsoft Office suite or power more advanced chatbots and digital assistants.
OpenAI's GPT-3, introduced in May, eclipses Turing-NLG with 175 billion parameters, easily making it the largest language model ever developed. A partial version of the tool was posted to GitHub in June.
A paper introducing GPT-3, written by more than two dozen OpenAI researchers and engineers, noted that the language model can write documents with a humanlike ability.
That, researchers noted, can have serious societal implications, especially if used by bad actors to spread false news reports.
Nick McQuireSenior vice president and head of AI and enterprise research, CCS Insight
It's unclear how Microsoft will use the model, but the tech giant could use GPT-3 to further build out its own language models, which power numerous products, including Microsoft Search, Project Cortex, and its Office suite.
"Whilst these massive models may be difficult to measure at this stage, the proof in the pudding is how this capability is brought into and enhances its products whether in SharePoint, PowerPoint and Office, Bing or Xbox, for example," McQuire said.
In a Sept. 22 blog post introducing the licensing deal, Kevin Scott, executive vice president and CTO at Microsoft, noted OpenAI will continue offering GPT-3 and other models through its Azure-hosted API.
"While we'll be hard at work utilizing the capabilities of GPT-3 in our own products, services and experiences to benefit our customers, we'll also continue to work with OpenAI to keep looking forward: leveraging and democratizing the power of their cutting-edge AI research as they continue on their mission to build safe artificial general intelligence," he said.