Microsoft is stepping up its game in the world of artificial intelligence by training a new, in-house AI language model to rival those from Google and OpenAI, according to a report from the Information. The new model, internally known as MAI-1, is being overseen by Mustafa Suleyman, a former Google DeepMind co-founder and CEO of AI startup Inflection. While the exact purpose of the model is still being determined, Microsoft may preview it at their upcoming Build developer conference.
MAI-1 is expected to be significantly larger than previous models Microsoft has trained, making it more expensive to develop. The company recently launched a smaller AI model called Phi-3-mini in an effort to attract a wider client base with cost-effective options. Microsoft has already invested billions in OpenAI and integrated the ChatGPT maker’s technology into its suite of productivity software, allowing it to take an early lead in the generative AI race.
To improve the MAI-1 model, Microsoft has set aside a large cluster of servers equipped with Nvidia GPUs and large amounts of data. The model is projected to have around 500 billion parameters, compared to OpenAI’s GPT-4 with one trillion parameters and Phi-3-mini with 3.8 billion parameters. Suleyman was appointed in March to lead Microsoft’s consumer AI unit and has brought in several employees from Inflection, although the new model is not directly linked to the startup.
As Microsoft continues to make advancements in AI technology, the company’s commitment to innovation and competition with industry giants like Google and OpenAI is evident. With the development of MAI-1 and the integration of AI technology across its suite of products, Microsoft is positioning itself as a leader in the fast-evolving AI landscape. The upcoming preview of MAI-1 at the Build conference will showcase the company’s dedication to pushing the boundaries of artificial intelligence and securing its foothold in the market.
