As reported by the Information on Monday, Microsoft is developing an AI language model internally that is big enough to take on rivals like Alphabet’s Google and OpenAI.
As per the report, which cited two Microsoft employees with knowledge of the effort, the newly hired Mustafa Suleyman, the former CEO of AI startup Inflection and co-founder of Google DeepMind, is in charge of the new model, internally called MAI-1.
The model’s precise goal is still unknown and will be based on how well it functions. The report stated that Microsoft might give a sneak peek at the new model at its Build developer conference later this month.
According to the report, MAI-1 will cost more because it will be “far larger” than the earlier, smaller, open-source models that Microsoft had trained beforehand.
Microsoft released Phi-3-mini, a scaled-down artificial intelligence model, last month in an effort to reach a larger customer base with more affordable choices.
The business has poured billions of dollars into OpenAI and integrated the technology of ChatGPT’s creator into its entire productivity software suite, giving it an early advantage in the race for generative AI.
According to the report, Microsoft has been allocating a sizable cluster of servers with Nvidia graphics processing units and a sizable amount of data to enhance the model.
According to the report, MAI-1 is expected to have approximately 500 billion parameters, whereas GPT-4 from OpenAI is said to have one trillion parameters, and Phi-3 mini measures 3.8 billion parameters.
In addition to hiring several Inflection staff members, Microsoft appointed Suleyman as the head of its recently established consumer AI division in March.
According to the report, the new model is not derived from Inflection, though it might use startup training data.