Microsoft reveals self-developed machine learning models during OpenAI discussions
Microsoft has taken a significant step forward in the world of artificial intelligence (AI) with the introduction of two new machine learning models: MAI-Voice-1 and MAI-1-preview.
The MAI-Voice-1 model, already powering Copilot Daily, an online AI-voiced summary of news and historic events, and Copilot Podcasts, is one of the most efficient speech systems available today. Capable of generating a full minute of audio in under a second on a single GPU, it stands out as a leader in the field.
MAI-1-preview, on the other hand, is being tested as the future basis of Microsoft's Copilot service. This in-house mixture-of-experts model, pre-trained and post-trained on approximately 15,000 NVIDIA H100 GPUs, is designed to provide powerful capabilities for consumers seeking models that specialize in following instructions and providing helpful responses to everyday queries.
Microsoft's collaboration with OpenAI has been successful so far, but the company's CEO, Mustafa Suleyman, stated that they need their own foundation models because AI is fundamental to their business. This move comes as Microsoft has invested around $13 billion in OpenAI and the two firms are trying to renegotiate their contract.
While OpenAI did not respond to a request for comment regarding Microsoft's new models, LMArena currently ranks Microsoft's MAI-1 model as the equal thirteenth most effective in terms of output quality. However, the MAI-1 model is not yet available to the public, with US would-be testers required to apply for access.
Meanwhile, Microsoft's GB200 cluster is now operational, and they expect to expose MAI-1-preview for specific Copilot scenarios in the coming weeks. It's unclear why Microsoft would continue paying for less capable technology if they can create the strongest models in the world.
Last year, Microsoft and OpenAI limited access to their speech synthesis projects due to potential abuses such as voice identification spoofing or impersonation. As the AI landscape continues to evolve, it will be interesting to see how these developments shape the future of AI technology.
