Microsoft is taking on the giants of the AI world with the development of its powerful new AI language model, MAI-1. With plans to rival Google and OpenAI, Microsoft is training this in-house model, which is said to be “far larger” than any previous models they’ve trained.
Award-winning AI researcher and co-founder of Google DeepMind, Mustafa Suleyman, is overseeing the development of MAI-1. Suleyman was also the former CEO of AI start-up Inflection, adding to his impressive track record in the field. The exact purpose of MAI-1 has not yet been determined and will depend on its performance.
While Microsoft declined to comment on the matter, reports suggest that they may provide a preview of MAI-1 at their upcoming Build developer conference later this month. This move could mark Microsoft’s entry into the competitive landscape of AI language models.
With a model consisting of roughly 500 billion parameters, MAI-1 is set to take on OpenAI’s GPT-4, which boasts one trillion parameters. Microsoft has made substantial investments in OpenAI, which has allowed them to integrate OpenAI’s ChatGPT technology into their suite of productivity software, giving them an early advantage in the generative AI race.
To ensure MAI-1’s success, Microsoft has dedicated large clusters of servers equipped with Nvidia’s graphic processing units and massive amounts of data to improve the model. This promising approach demonstrates Microsoft’s commitment to making significant advancements in the AI space.
Suleyman’s appointment as the head of Microsoft’s newly created consumer AI unit and the hiring of several Inflection employees further emphasize the company’s dedication to AI development. Although MAI-1 is a separate project, there is potential for it to benefit from Inflection’s training data.
As the race for AI dominance heats up, Microsoft’s entry into the field will undoubtedly add a new level of competition. With the potential preview of MAI-1 at the Build developer conference, it will be fascinating to see how Microsoft’s new AI language model stacks up against the likes of Google and OpenAI.
Use the share button below if you liked it.