Microsoft and OpenAI have had a long-standing relationship where the former powers a suite of AI tools using OpenAI’s AI models. However, Microsoft is reportedly ready to branch out and make AI models of its own.
A report by The Information hints that Microsoft is developing an LLM called “MAI-1”. It is being speculated that the “M” stands for “Microsoft” and the “1” suggests that this might be the AI’s first iteration. The project is being led by Mustafa Suleyman, co-founder of Google’s DeepMind. This is the largest AI model Microsoft will develop in-house since it invested more than $10 billion in OpenAI in exchange for the right to utilize the startup’s models. Microsoft Copilot is also powered by OpenAI’s GPT-4.
MAI-1 will require more processing power and training data than Microsoft’s previous open-source models, like Phi-3, because it will have over 500 billion parameters. According to reports, this puts MAI-1 well above smaller models like Meta and Mistral’s 70 billion parameter models and in a similar league as OpenAI’s GPT-4, which is reported to have over 1 trillion parameters.
The development of MAI-1 points to a dual strategy for AI at Microsoft, with an emphasis on both small, locally run language models for mobile devices and more advanced, cloud-powered models.
One of The Information’s sources claims that the exact purpose of MAI-1 is still not known (not even within Microsoft) and that its most effective application will be based on how well it performs. Microsoft has been assigning a sizable cluster of servers with Nvidia GPUs to train the model, and it has been gathering training data from multiple sources, such as text produced by OpenAI’s GPT-4 and publicly available Internet data.
The Information claims that Microsoft may showcase MAI-1 at its Build developer conference later this month, depending on the progress made in the upcoming weeks.
It’s unclear how this will impact the working relationship between Microsoft and OpenAI. Will Microsoft continue to let OpenAI power Copilot, or will its own AI take over? We’ll have to wait and see what, if anything, Microsoft is working on.
- The Wait Is Over: OpenAI Finally Introduced Sora - December 12, 2024
- Google’s AI Video Generator Launches in Private Preview, Ahead of OpenAI’s Sora - December 11, 2024
- Meet Nova: Amazon’s New Family of Multimodal AI Models - December 10, 2024
