Microsoft is reportedly making a significant shift in its AI strategy by developing an in-house large language model (LLM) named MAI-1. This move marks a departure from the company’s previous focus on collaborating with OpenAI and utilizing its models, including the GPT-4 family of AIs, to power various Microsoft Copilot products.
The decision to create MAI-1 suggests that Microsoft aims to exert more control over its technology, especially following the hiring of Mustafa Suleyman, a former Google AI chief and the founder of the Inflection AI lab. While Suleyman and his team have joined Microsoft, MAI-1 is expected to be a completely new model.
According to a report by The Information, Microsoft’s substantial investment of over $10 billion in OpenAI for AI model rights may lead to GPT-4 no longer being the engine behind Copilot, the consumer-oriented chatbot available on the Windows 11 desktop.
The development of two large language models concurrently could indicate Microsoft’s experimentation with different approaches. Chatbot interfaces have gained popularity but incur high operational costs due to the substantial data transfer involved. Moreover, concerns over query privacy arise when data is sent to external servers for processing.
While several companies, like Apple for its upcoming iOS 18 release, are exploring on-device data processing to address these concerns, the computational requirements of large language models present a challenge. By pursuing both a smaller-scale LLM and a more extensive model simultaneously, Microsoft seeks to offer a diverse range of solutions to users.
