Microsoft has a lot Nvidia Graphics processing units, but they do not use them to develop state -of -the -art models for artificial intelligence.
There are good reasons for this position, said Mustafa Sulyman, the CEO of the AI company, in an interview on Friday to Steve Kovach from CNBC. Waiting for the construction of models that are “three or six months behind” offers several advantages, including lower costs and the ability to concentrate on certain applications, said Sulyman.
It is “cheaper to give a certain answer as soon as you have waited for the first three or six months until the border goes first. We call this from Frontier,” he said. “This is actually our strategy to really play a very close second in view of the capital intensity of these models.”
Sulyman made a name for himself as a co-founder of Deepmind, the AI laboratory that Google bought in 2014, reports $ 400 to $ 650 million. Sulyman arrived in Microsoft with other employees last year, where he was CEO.
Microsoft more than ever counts on relationships with other companies to grow.
It receives AI models from San Francisco Startup Openaai and additional computing power from new public Coreweave in New Jersey. Microsoft has repeatedly enriched Bing, Windows and other products with the latest Openai systems to write human language and generate pictures.
Microsoft's Copilot will win “Speicher” to keep important facts about people who repeatedly use the assistant, Sulyman said on Friday at an event in Microsoft's Redmond, Washington, headquarters to remind the company's 50th birthday. This function was first opened to open Chatgpt from Openai with 500 million weekly users.
Via chatt, people can access first-class large language models such as the O1 argumentation model that takes time before spitting an answer. Openaai presented this ability in September – only weeks later Microsoft brought a similar ability called Think deeper to copilot.
Microsoft occasionally publishes open source models with a small language that can be carried out on PCs. They do not require a powerful server GPUS, which makes them differ from the O1 from Openaai.
Openai and Microsoft held a close relationship shortly after the startup started at the end of 2022, which effectively begins the generative AI race. In total, Microsoft has invested $ 13.75 billion in the startup, but recently the cracks have started in the relationship between the two companies.
Microsoft added to his list of competitors in July 2024, and Openaai announced in January to work with the competing cloud provider oracle For the 500 billion dollar -Stargate project. This came after years of Openai, which rely exclusively on the Azure Cloud from Microsoft. Despite the Openai, who works with Oracle, Microsoft announced in a blog post that the startup “recently received a new, large azure obligation”.
“Look, it is absolutely mission -critical that we can agree with Microsoft ourselves,” said Sulyman. “At the same time, I think about these things over five and ten years. You know that by 2030 we are at least deeply worked with Openai who have [had an] Extremely successful relationship for us.
Microsoft focuses on internal AI, but the company is not urging itself to build up the most modern models, said Sulyman.
“We have an incredibly strong AI team, huge amounts of calculation, and it is very important to us that we, perhaps not the absolute border, do not develop the best model in the world at first,” he said. “This is very, very expensive and unnecessary to cause this duplication.”
REGARD: Microsoft Copilot The beginning of a seismic shift in AI integration, says Microsoft Ai -CEO Sulyman
Comments are closed.