Microsoft in battle with OpenAI?

Despite its partnership with OpenAI, Microsoft is creating its own in-house large language model (LLM) called MAI-1, which is expected to compete with (and beat) ChatGPT and could be revealed at its Build developer conference at the end of May.

How will MAI-1 compare with GPT and other LLMs?

MAI-1 will utilize approximately 500B parameters, which makes it a direct rival to other industry-leading LLMs, aside from GPT, such as Google’s Gemini, and Anthropics Claude.    

It’s expected that MAI-1 will have similar capabilities as OpenAI’s ChatGPT-4 and will surpass its newest small language model, Phi-3, which is suitable for on-device use cases, in terms of size and cost.

Although it seems like Microsoft is competing with its partner, OpenAI, this is clearly a strategic move, as its investment will allow it to integrate and benefit from the GPT technology into the new LLM, giving it a competitive edge in the AI field.

Who is leading the development of MAI-1?

Microsoft's new consumer AI unit, led by Google DeepMind co-founder and ex-Inflection CEO, Mustafa Suleyman, will spearhead the development of MAI-1. Rumor has it that Microsoft has access to Infelections technology, so although the new model won’t be a continuation of Inflection's work, it may at least utilize the start-up's training data.