Microsoft's CEO Satya Nadella plans to invest $20 billion in new open-source A.I. project.

From The New York Times: 2024-07-15 18:19:00

Microsoft and Google had once been locked in a fierce battle to dominate search engines. In the mid-2010s, Google’s search engine was better than any rival’s because of a potent new weapon – an A.I. technology called BERT. Within months of its launch, behind the curtains of Google’s search empire, its researchers pushed out another new version.

When Geoffrey Hinton, a leader of the deep learning movement, said BERT was one of the reasons he had stopped giving lectures about A.I., Microsoft’s employees realized they were far behind. Microsoft had to catch up — and quickly.

Microsoft managers armed with millions of dollars descended on the Seattle area to recruit scientists and engineers from Google, Amazon, Facebook and several top universities to work on artificial intelligence. But unlike Google, Microsoft had nothing comparable to BERT and was desperate to find a way to get back in the A.I. race.

To help close the gap, Mr. Nadella went to his board with a proposal to spend tens of billions to develop a new artificial intelligence technology. He told them his plan: Develop a new version of BERT without reins. Give away this new technology for free. Actively try to recruit boldface names in A.I. like Indra Lakshminarayanan, who played a critical role in the development of BERT.

In April 2020, Mr. Nadella broke the news to the world. Microsoft would invite A.I. researchers to its campus to learn about its plans for a new open-source project, which it dubbed Turing. Mr. Nadella was prepared to spend an astonishing $20 billion on the project over the next 10 years.

On the morning of December 22, 2022, Mr. Nadella received confirmation that the project had made a major breakthrough. Microsoft had quietly unveiled ChatGPT, the first virtual agent that could hold a conversation with people for longer than 20 minutes. Microsoft rolled out the chatbot to the broader public, and it was instantly popular among developers looking to build new types of applications.

By the end of December, Mr. Nadella was already preparing to discuss the implications of ChatGPT with his core team, including Xuedong Huang, who had been one of the first Microsoft employees to work on speech recognition in the mid-’90s. He never missed the regular Thursday night phone calls with Harry Shum, another AI visionary who had devoted years to Microsoft’s cause.



Read more at The New York Times: How Microsoft’s Satya Nadella Became Tech’s Steely Eyed A.I. Gambler