Microsoft AI launches its first in-house models

Microsoft’s AI division announced its first AI models on Thursday. The company claims that its new Mai-Voice-1 speech model can generate a minute of audio in less than a second on a single GPU, while May-1-Propew “offers an overview of future offers inside Copilot”.

You can try Ma1-Voice-1 for yourself on Copilot Labs, where you can enter what you want the IA model, as well as to change his voice and his style of speech. In addition to this model, Microsoft introduced Mai-1-Preview, which it indicates that it was formed out of approximately 15,000 Nvidia H100 GPU. It is designed for users who need an AI model capable of following instructions and “providing useful responses to daily queries”.

Microsoft AI chief, Mustafa Suleyman, said during an episode of Decoder Last year, the company’s internal AI models are not focused on business use cases. “My logic is that we have to create something that works extremely well for the consumer and really optimize for our use case,” said Suleyman. “So, we have large amounts of very predictive and very useful data on the side of advertising, consumption telemetry, etc.

Microsoft AI plans to deploy Mai-1-Preview for certain cases of use of text in its Assistant Copilot AI, which is currently based on the large Openai language models. He also started to publicly test his Mai-1-Preview model on the comparative analysis platform of the LMARENA.

“We have great ambitions for where we are going,” writes Microsoft AI in the blog post. “Not only will we continue other advances here, but we think that the orchestration of a range of specialized models serving different user intentions and use cases will unlock a huge value.”