Microsoft celebrates 50 years with Copilot

Microsoft’s AI division head, Mustafa Suleyman, believes there are benefits to using AI models that are a few months behind the cutting edge. Building the most advanced models is expensive and can be duplicative. “It’s absolutely mission-critical that long-term, we are able to do AI self-sufficiently at Microsoft,” Suleyman said.

However, the company is not currently using its resources to develop state-of-the-art AI models.

Waiting to build models that are “three or six months behind” offers advantages such as lower costs and the ability to focus on specific use cases. “That’s actually our strategy, to really play a very tight second, given the capital-intensiveness of these models,” Suleyman explained.

Microsoft relies on partnerships with other companies to grow. It sources AI models from OpenAI and computing power from a company in New Jersey.

Microsoft’s AI strategy and Copilot

Microsoft has integrated OpenAI’s systems into products like Bing and Windows. Suleyman announced updates to Microsoft’s Copilot assistant at an event on Friday.

Copilot will now be able to remember key facts about people who use it repeatedly, a feature first seen in OpenAI’s ChatGPT.

See also  US consumer prices rise sharply in January

Microsoft occasionally releases open-source models that can run on PCs without requiring powerful server GPUs. These differ from OpenAI’s cutting-edge offerings. Despite some recent issues, Suleyman stressed the importance of Microsoft being able to do AI independently in the long run.

However, he acknowledged that Microsoft is not currently pushing to build the most advanced models. “We have an incredibly strong AI team, huge amounts of compute, and it’s very important to us that we don’t necessarily develop the absolute frontier, the best model in the world first,” he said. “That’s very, very expensive to do and unnecessary to cause that duplication.”

Kayla Lear

More Stories