Microsoft I own a lot nvidia Although it is a graphics processing unit, it does not use them to develop cutting-edge artificial intelligence models.
There is good reason for that position, AI CEO Mustafa Suleyman told CNBC’s Steve Kovach in an interview Friday. Waiting to build a “three or six months behind” model offers several benefits, including cost reductions and the ability to focus on specific use cases, Suleyman said.
“If you wait for the first three or six months for the frontier to go first, it’s cheaper to give a specific answer. We call it off frontier,” he said. “It’s actually our strategy. Given the capital strength of these models, it’s actually playing very tight seconds.”
Suleyman created his own name for $400 million to $650 million as co-founder of the AI ​​Lab, which Google purchased in 2014. Suleyman arrived at Microsoft last year along with other employees of Startup’s fluctuations.
More than ever, Microsoft expects to grow its relationships with other companies.
Get AI models from San Francisco startup Openai and get newly released supplemental computing power coreweave In New Jersey. Microsoft repeatedly concentrates Bing, Windows and other products, writing human-like languages ​​with Openai’s latest system and generating images.
Microsoft co-pilots will gain “memories” to hold important facts about people who use their assistants repeatedly, Suleyman said Friday in commemoration of the company’s 50th birthday at an event at its headquarters in Redmond, Microsoft, Washington. That feature first appeared in Openai’s ChatGPT.
Through CHATGPT, people can access large language models of top flights, such as the O1 inference model, which takes time before spitting out the answer. Openai introduced its feature in September. A few weeks later, Microsoft brought a similar feature called Think Deeper to Copilot.
Microsoft may release open source small language models that can be run on PCs. They don’t need a strong server GPU, which is different from Openai’s O1.
Openai and Microsoft formed a tough relationship shortly after the startup launched their ChatGPT chatbot in late 2022, effectively starting a generation AI race. In total, Microsoft has invested $13.75 billion in startups, but recently there has been a rift in relations between the two companies.
Microsoft added Openai to its competitor list in July 2024, and announced in January that Openai was working with rival cloud providers Oracle About the $500 billion Stargate project. It came after many years of Openai, relying solely on Microsoft’s Azure Cloud. Despite Openai’s partnership with Oracle, Microsoft announced in a blog post that the startup has “recently made a new, massive Azure commitment.”
“Look, in the long run, we’re absolutely critical of the mission and we can do AI self-sufficiently at Microsoft,” Suleyman said. “At the same time, I think of these things for five and ten years. At least until 2030, we are working deeply with Openai, who has a very successful relationship for us.
While Microsoft focuses on building its own AI internally, the company has not pushed itself to build its most cutting-edge models, Suleyman said.
“We have a very strong AI team and we have a huge amount of calculations. It’s very important to us. We may not develop an absolute frontier, perhaps the best model in the world,” he said. “It’s very expensive and doesn’t have to cause that overlap.”
Watch: The start of the earthquake shift for Microsoft Copilot AI integration, says Microsoft AI CEO Suleyman
