Microsoft owns a lot of Nvidia graphics processing units, nevertheless it is not using them to develop state-of-the-art artificial intelligence models.
There are good reasons for that position, Mustafa Suleyman, the corporate’s CEO of AI, told CNBC’s Steve Kovach in an interview on Friday. Waiting to construct models which can be “three or six months behind” offers several benefits, including lower costs and the flexibility to focus on specific use cases, Suleyman said.
It’s “cheaper to present a selected answer once you have waited for the primary three or six months for the frontier to go first. We call that off-frontier,” he said. “That is actually our strategy, is to essentially play a really tight second, given the capital-intensiveness of those models.”
Suleyman made a reputation for himself as a co-founder of DeepMind, the AI lab that Google bought in 2014, reportedly for $400 million to $650 million. Suleyman arrived at Microsoft last 12 months alongside other employees of the startup Inflection, where he had been CEO.
Greater than ever, Microsoft counts on relationships with other corporations to grow.
It gets AI models from San Francisco startup OpenAI and supplemental computing power from newly public CoreWeave in Recent Jersey. Microsoft has repeatedly enriched Bing, Windows and other products with OpenAI’s latest systems for writing human-like language and generating images.
Microsoft’s Copilot will gain “memory” to retain key facts about individuals who repeatedly use the assistant, Suleyman said Friday at an event in Microsoft’s Redmond, Washington, headquarters to commemorate the corporate’s fiftieth birthday. That feature got here first to OpenAI’s ChatGPT, which has 500 million weekly users.
Through ChatGPT, people can access top-flight large language models corresponding to the o1 reasoning model that takes time before spitting out a solution. OpenAI introduced that capability in September — only weeks later did Microsoft bring a similar capability called Think Deeper to Copilot.
Microsoft occasionally releases open-source small-language models that may run on PCs. They do not require powerful server GPUs, making them different from OpenAI’s o1.
OpenAI and Microsoft have held a good relationship shortly after the startup launched its ChatGPT chatbot in late 2022, effectively kicking off the generative AI race. In total, Microsoft has invested $13.75 billion within the startup, but more recently, fissures in the connection between the 2 corporations have begun to point out.
Microsoft added OpenAI to its list of competitors in July 2024, and OpenAI in January announced that it was working with rival cloud provider Oracle on the $500 billion Stargate project. That got here after years of OpenAI exclusively counting on Microsoft’s Azure cloud. Despite OpenAI partnering with Oracle, Microsoft in a blog post announced that the startup had “recently made a brand new, large Azure commitment.”
“Look, it’s absolutely mission-critical that long-term, we’re capable of do AI self-sufficiently at Microsoft,” Suleyman said. “At the identical time, I take into consideration this stuff over five and 10 12 months periods. You realize, until 2030 at the least, we’re deeply partnered with OpenAI, who’ve [had an] enormously successful relationship for us.
Microsoft is concentrated on constructing its own AI internally, but the corporate shouldn’t be pushing itself to construct probably the most cutting-edge models, Suleyman said.
“We’ve got an incredibly strong AI team, huge amounts of compute, and it’s extremely essential to us that, you already know, perhaps we do not develop absolutely the frontier, the most effective model on the earth first,” he said. “That is very, very expensive to do and unnecessary to cause that duplication.”
WATCH: Microsoft Copilot starting of a seismic shift in AI integration, says Microsoft AI CEO Suleyman