• Source:Reuters
Add as a preferred

Tech giant Microsoft has grabbed regular headlines for integrating artificial intelligence into its existing and new products. The company is now reportedly training a large language model (LLM) to take on Google, OpenAI and more AI-facing companies, according to The Information. Learned to be 'MAI-1', the project is undergoing development under the leadership of the newly hired Mustafa Suleyman, Google DeepMind co-founder and former CEO of Inflection.

While the exact intent of the AI model is unclear at the moment, the use cases may depend on its performance. The report noted that the company could showcase the product later this month at its Build developer conference. MAI-1 is said to be larger than the earlier models the company has trained. This may imply it may come at a premium price.

Also Read: Microsoft Introduces New Lightweight AI Model That Can Run On Smartphones

For the unaware, the company also unveiled a small AI model Phi-3-mini, to draw more customers and introduce a cost-effective alternative. In the past, the company had infused massive amounts of cash (over $10 billion) in OpenAI to introduce the viral chatbot ChatGPT's tech into its offerings. This added an advantage for the company in the AI race.

The Information said that the company is also allocating servers with Nvidia GPUs and large data clusters for the new initiative. It could incorporate 500 billion parameters, compared to GPT 4's one trillion and Phi-3-mini's 3.8 billion. As for the new leadership, the company welcomed Suleyman in March along with several other Inflection employees.

Also Read: Why Microsoft's New London-Based AI Hub Is Important?

The model could be trained using the data from the startup but not be a carried-over product from the AI company, according to Microsoft employees cited in the report. These latest developments around MAI-1 hint the company plans to introduce a host of AI models ranging from small-scale offerings for smartphones that can run locally to large-scale flagship models that may depend on cloud connectivity.