Microsoft’s AI division has unveiled its first proprietary AI models: MAI-Voice-1 and MAI-1-preview. The MAI-Voice-1 speech model can generate a minute of audio in under one second using just one GPU, while MAI-1-preview, trained on around 15,000 Nvidia H100 GPUs, aims to enhance user interaction by efficiently responding to everyday queries. Available on Copilot Labs, users can customize the MAI-Voice-1 model’s voice and style. Microsoft AI chief Mustafa Suleyman emphasized a consumer-centric approach, focusing on developing models that deliver practical benefits based on predictive data. The MAI-1-preview will be integrated into Copilot AI, which currently utilizes OpenAI’s models, and is undergoing public tests on the LMArena benchmarking platform. Microsoft’s future vision includes launching a diverse range of specialized AI models to create significant value across various user intents and use cases, setting ambitious goals for ongoing advancements.
Source link
