Microsoft is developing a new in-house AI language model called MAI-1, aiming to compete with models from Google and OpenAI, according to a report by The Information. The project is being led by Mustafa Suleyman, co-founder of Google DeepMind and former CEO of AI startup Inflection. The exact purpose of the model is yet to be determined and will depend on its performance.
MAI-1 is expected to be "far larger" and more expensive than Microsoft's previous open-source models. Last month, Microsoft launched a smaller AI model, Phi-3-mini, to attract a broader client base with cost-effective options. Microsoft has invested heavily in OpenAI and integrated the ChatGPT maker's technology across its productivity software suite, gaining an early advantage in the generative AI race.
To improve the model, Microsoft has been allocating large server clusters equipped with Nvidia's graphic processing units and large data volumes. The MAI-1 model is expected to have about 500 billion parameters, compared to OpenAI's GPT-4's one trillion parameters and Phi-3 mini's 3.8 billion parameters.
In March, Microsoft appointed Suleyman as the head of its newly formed consumer AI unit and hired several Inflection employees. The new model is not a continuation of Inflection's work, but it may use training data from the startup.
