Anyone in the AI industry knows about Microsoft’s and OpenAI’s symbiotic relationship. The former uses the latter’s AI models to power its suite of AI tools. That does not mean that Microsoft does not have AI models of its own. According to the report, Microsoft is working on a new LLM named MIA-1, and it’s meant to compete with the best models on the market.
Microsoft, just like Google, is a very far-reaching company. As such, both companies have introduced several language models for different purposes. For example, Google’s Gemma model is small enough to fit on a computer. Also, Microsoft’s Phi-3 model is also a bit smaller in scale and very specialized.
Microsoft could be working on a large LLM named MIA-1
Microsoft already has a ton of power at its fingertips with OpenAI’s models. However, Microsoft is a big boy; it can make its own LLM. The company is looking to make a model that’s big enough to compete with the best models on the market.
Named MIA-1, this model is said to potentially have up to 500 billion parameters. For comparison, Phi-3 Mini has 3.8 billion parameters. So, this is going to be much larger in scale.
We don’t know much more about this model’s capabilities. However, we know that Microsoft got one of the brightest minds in AI to lead the development. The company got Mustafa Suleyman to oversee the development of MIA-1. Suleyman is one of the co-founders of Google Deepmind, and he is a former CEO of Inflection, an AI startup.
Microsoft has been developing large servers with a ton of Nvidia GPUs to help build this model. We don’t know how many, but we know that it’s a substantial amount. Right now, information is still scarce about MIA-1. We’re not certain if the company is planning on using this as a replacement for OpenAI’s models. If so, then what will happen to their partnership? We will just have to find out as this story unfolds.
