Microsoft is training a fresh in-house artificial intelligence language model immense enough to compete with the languages of Google and Alphabet’s OpenAI, it said Monday.
The fresh model, internally called MAI-1, is being overseen by recently hired Mustafa Suleyman, co-founder of Google DeepMind and former CEO of AI startup Inflection, the report said, citing two Microsoft employees with knowledge of the effort.
The exact purpose of the model has not yet been determined and will depend on how well it performs. The report suggests that Microsoft will be able to showcase the fresh model as early as this month at the Build developer conference.
Microsoft declined to comment when contacted by Reuters.
The report says MAI-1 will be “significantly larger” than previous, smaller open source models that Microsoft has previously trained, which means it will be more pricey.
Last month, Microsoft launched a smaller AI model called Phi-3-mini that aims to appeal to a broader customer base with cost-effective options.
The company has invested billions of dollars in OpenAI and implemented the manufacturer’s ChatGPT technology in its productivity software suite, giving it an early leadership position in the generative AI race.
The report shows that Microsoft has dedicated a immense cluster of servers equipped with Nvidia GPUs along with a immense amount of data to improve the model.
According to the report, MAI-1 will have approximately 500 billion parameters, while OpenAI’s GPT-4 will have one trillion parameters and Phi-3 mini will measure 3.8 billion parameters.
In March, Microsoft named Sulejman to head its newly created consumer artificial intelligence unit and hired several Inflection employees.
The fresh model is not ported from Inflection, although it may be based on training data acquired since launch, the report added.