LIVE UPDATES
Microsoft is developing a major new in-house artificial intelligence (AI) language model to compete with the industry leaders from Google and OpenAI, according to a report from The Information.
The model, codenamed MAI-1, is under the direction of Mustafa Suleyman, former CEO of Inflection and co-founder of Google DeepMind.
Microsoft sources cited by The Information said the purpose of MAI-1 is still being refined and will be performance-dependent, adding that the tech giant may offer a preview of MAI-1 at its Build developer conference later in May.
However, Microsoft has refused to publicly comment on the development.
According to The Information, MAI-1 will be significantly larger and thus more costly to operate than prior open-source Microsoft AI models.
The company's substantial investment in OpenAI and its utilization of ChatGPT technology across its products positions it well in the burgeoning field of generative AI.
MAI-1 development involves extensive use of Nvidia graphic processing units and massive amounts of data for training. The Information suggests it will have approximately 500 billion parameters, considerable but smaller than OpenAI's reported one trillion for GPT-4.
Microsoft formed a consumer AI unit led by Suleyman in March, additionally hiring a number of Inflection employees. Though separate from Inflection's work, MAI-1 may leverage some of the startup's training data.
