Satya Nadella: DeepSeek’s R1 First AI Mannequin to Come Near OpenAI’s

Microsoft CEO Satya Nadella stated the Chinese language AI startup DeepSeek’s R1 mannequin was the primary AI mannequin he had seen that was as aggressive as OpenAI’s.

“OpenAI has been to date forward that nobody’s actually come shut,” Nadella stated in an interview with Bloomberg Businessweek printed Thursday.

“DeepSeek, and R1 specifically, was the primary mannequin I’ve seen publish some factors,” Nadella added.

In January, DeepSeek spooked buyers after its free chatbot app hit the highest spot on Apple’s US App Retailer’s rating. DeepSeek’s high-performing however comparatively cheaper fashions sparked a sell-off in AI-related shares as buyers questioned if it might trigger demand for AI {hardware} like Nvidia’s chips to fall.

Microsoft began providing variations of DeepSeek’s R1 mannequin on its cloud platform, Azure AI Foundry, in January. Other than DeepSeek, the platform consists of different AI fashions from firms like OpenAI, Meta, and Mistral.

Utilizing R1 on Microsoft’s platform meant that knowledge wouldn’t be despatched to DeepSeek’s servers in China.

Asha Sharma, a Microsoft company vice chairman who heads the corporate’s AI platform product, wrote in a weblog publish in January that R1 went by way of “rigorous pink teaming and security evaluations” earlier than it was made obtainable to clients.

In January, OpenAI CEO Sam Altman known as R1 “a powerful mannequin, notably round what they’re capable of ship for the value.” Altman stated OpenAI would “pull up some releases” in response to DeepSeek’s “invigorating” competitors.

“R1 is definitely not that uncommon,” Ben Buchanan, a former particular advisor for synthetic intelligence within the Biden administration, stated in an episode of “The Ezra Klein Present” which aired in March.

Buchanan stated whereas DeepSeek’s engineers are “extraordinarily proficient,” he did not suppose the “media hype round it was warranted.”

“The place do you suppose they obtained their efficiency will increase from? We learn their papers. They’re good people who find themselves doing precisely the identical form of algorithmic effectivity work that firms like Google and Anthropic and OpenAI are doing,” Buchanan added.

Microsoft, OpenAI, and DeepSeek didn’t reply to requests for remark from Enterprise Insider.