May 07, 2024
Microsoft, an American multinational technology company, is creating a large-scale artificial intelligence (AI) language model to compete with Google and OpenAI.
Citing two Microsoft workers familiar with the project, The Information reported that Mustafa Suleyman, co-founder of Google DeepMind and previous CEO of AI firm Inflection, is overseeing the model.
The tech giant may unveil the new model at its Build developer conference later this month. However, the model's function is unknown and will be determined based on its performance.
In addition, MAI-1 would be "far larger" than previous smaller, open-source models trained by Microsoft, making the new model more expensive.
Last month, the Big Tech business introduced a smaller AI model called Phi-3-mini to reach a larger client base with more cost-effective solutions.
Moreover, the company has spent billions of dollars on OpenAI and integrated the ChatGPT maker's technology into its suite of productivity products, giving it an early advantage in the generative AI race.
According to the article, Microsoft has allocated a big cluster of computers equipped with Nvidia's graphic processing units, along with enormous quantities of data, to improve the model.
MAI-1 will include around 500 billion parameters, vs one trillion for OpenAI's GPT-4 and 3.8 billion for Phi-3 mini.
