On Thursday, Microsoft announced its first in-house AI models, marking a significant step towards greater independence from OpenAI. The introduction of MAI-Voice-1 AI and MAI-1-preview signifies a strategic shift in the complex collaboration between the two tech giants.
Advanced speech and language technology
The new MAI-Voice-1 speech model can generate a minute of audio in just one second on a single GPU. Microsoft is already using this technology for Copilot Daily, where an AI host reads daily news stories, and for creating podcast-like discussions to clarify topics.
The MAI-1 preview model was trained on approximately 15,000 Nvidia H100 GPUs and focuses on following instructions and providing useful answers to everyday questions. Users can already try out the speech model through Copilot Labs, where they can input text and customize the voice and speaking style.
Strategic shift towards the consumer market
Mustafa Suleyman, head of Microsoft AI, emphasized last year that the internal AI models are not primarily aimed at business customers. "We need to create something that works excellently for the consumer and truly optimize for our use," he stated. Microsoft has extensive data on advertisements and consumer behavior, which provides an advantage in developing consumer-focused AI assistants.
Impact for companies and organizations
This development has significant implications for the AI market. Companies relying on Microsoft's Copilot services will gain access to new capabilities without increased dependence on external AI providers like OpenAI. This could lead to more stable pricing and greater control over AI functionalities.
Future Perspectives
Microsoft plans to gradually roll out MAI-1-preview in Copilot for text usage, which currently still relies on OpenAI's language models. The company is also publicly testing the model on the AI benchmark platform LMArena. "We have big ambitions for the next steps," said Microsoft AI. The company believes that orchestrating specialized models for different use cases will create significant value.
This strategic move positions Microsoft as a direct competitor to OpenAI and other AI developers, with its own technology at the forefront instead of partnerships.
The Verge