Microsoft is making ‘significant investments’ in training its own AI models
Microsoft AI launched its first internal models last month, adding to the already complicated relationship with its partner Openai. Now, Microsoft AI, Mustafa Suleyman, said that the company makes “important investments” in the required calculation capacity for Microsoft’s own frontal models.
“We should have the ability to build world -class border models in the house of all sizes, but we must be very pragmatic and use other models where we need it,” Suleyman said at the town hall only of Microsoft employees. “We are also going to make significant investments in our own cluster, so today May-1-preview has only been formed on 15,000 h100, a tiny cluster in the big scheme of things.”
Suleyman suggested that Microsoft has ambitions to form models comparable to Meta, Google and Xai’s efforts on clusters that are “six to ten times larger” than Microsoft used for its May-1 forecast. “Much more to do, but it’s good to take the first steps,” said Suleyman.
Microsoft’s CEO Satya Nadella said in the same town hall that he “impatiently awaited the construction model capacity of us, so that we can build products in model mode”. Nadella has also clearly indicated that Microsoft “would definitely support several models” in her products and highlighted the Github co -pilot as “the best example” of this strategy.
Microsoft would have planned to use anthropic AI models soon for certain Microsoft 365 features. Information The reports that the Microsoft 365 co -pilot will be “partially fueled by anthropogenic models”, after Microsoft discovered that some of these models have outperformed Openai in Excel and PowerPoint.
