On Independence Day, Microsoft announced the launch of MAI-1-preview, a cutting-edge AI model utilizing a mixture-of-experts architecture, pre-trained on 15,000 Nvidia H100 GPUs. This infrastructure is notably smaller than competitors’ 100,000 GPU clusters, yet Microsoft aims to enhance performance by deploying MAI-1-preview on Nvidia’s advanced GB200 cluster. This development is pivotal for Microsoft, marking a shift from its previous reliance on OpenAI’s models, after investing around $13 billion since 2019. The move is particularly significant as Microsoft transitions to independent AI model development by hiring Mustafa Suleyman, co-founder of DeepMind, along with his Inflection AI team. This strategic initiative is expected to bolster Microsoft’s AI capabilities, especially in its Copilot products, while Inflection AI continues to innovate under its new CEO. As the AI landscape evolves, Microsoft’s focus on independence and performance could reshape industry standards.
Source link
Share
Read more