“Microsoft” has announced the launch of the first artificial intelligence models internally, and the first model bears the name “Mai-Voice-1”, which is dedicated to generating natural sounds, while the second model comes under the name “Mai-1-PREVIEW” and is classified as a textual basis model that has been fully developed within the company.
According to the company’s statement, the Mai-Voice-1 model is characterized by its ability to generate a full accurate sound in less than one second using only one graphics processing unit (GPU), and the model is already used in some “Copolot” services such as Copilot Daily, which provides a daily audio summary of news, in addition to the production of podcast similar discussions to clarify topics, as can be possible Users have his experience via the Copilot Labs platform with the ability to adjust the sound tone and the style of dumping.
The MAI-1-PREVIEW text model has been trained using about 15,000 NVIDIA H100 chips, and is designed to deal with text instructions and provide useful responses for daily inquiries.
“Microsoft” confirmed that this model provides a glimpse of what will be presented in the future within the “Cobilot” system, as the company has already started testing it via the LMARNA platform for measuring the performance of artificial intelligence models, to be gradually integrated into some “Cobyot” services in the coming weeks.
The move comes at a time when the services of the “Cobilot” are still dependent mainly on the technologies of Openai, and Microsoft sees that the combination of a group of specialized models that serve various needs and uses will provide a huge value to users, and paves the way for a new stage of competition in the global artificial intelligence race.
