Microsoft developing new in-house AI model MAI-1, CTO clarifies not competing with OpenAI
As technology companies race to develop more advanced artificial intelligence, Microsoft is making strides with its own in-house model. According to reports, the tech giant is currently training a new large language model called MAI-1 that could rival cutting-edge AI like OpenAI's GPT-4.
Sources indicate MAI-1 will feature around 500 billion parameters, putting it on par with models from Google, Meta and Anthropic in terms of scale and capabilities. Developing such a sophisticated AI model requires immense computing power and data resources. Microsoft is said to be harnessing Nvidia GPUs and dedicating significant infrastructure to properly train MAI-1.
While some speculated this marked Microsoft entering direct competition with partner OpenAI, the company's CTO Kevin Scott moved quickly to clarify the nature of their relationship remains collaborative. In a LinkedIn post, Scott explained Microsoft and OpenAI have a proven track record building progressively more powerful supercomputers that fuel the next generation of AI advances.
Rather than competing, Scott emphasized both organizations intend to keep pushing each other forward for the greater benefit of the field. Microsoft will continue crafting state-of-the-art hardware while OpenAI leverages these resources to develop cutting-edge models available through products and services.
For their part, Microsoft researchers have also been working on smaller and larger AI models independently. However, the primary goal stated is collaboration, not competition, as the two powerhouses work in tandem toward new frontiers in artificial intelligence. As details on MAI-1's full capabilities remain undisclosed, its development marks an exciting step for Microsoft and AI overall.
