In 2019, Microsoft's highest-level executives were exchanging emails anxiously about AI, which they would eventually start investing in.
At the time, Chief Technology Officer Kevin Scott sent a four-page email to CEO Satya Nadella and co-founder Bill Gates with the subject line “Thoughts on OpenAI,” in which he wrote: He outlined concerns that Microsoft is falling far behind in the AI race. At that time, it was still more than four years before ChatGPT was released to the public. But Scott still found that OpenAI and Google had made extraordinary progress in his AI efforts.
“What's interesting about the efforts of OpenAI, DeepMind, and Google Brain is the scale of their ambitions,” Scott wrote.
The heavily redacted email was revealed as part of the Justice Department's antitrust investigation into Google, and was first reported by Business Insider.
That same year, Microsoft will invest $1 billion in OpenAI, the very company Scott mentioned in his email. Ultimately, the tech giant would go on to invest at least another $10 billion in the startup, and with the launch of ChatGPT, he is credited with popularizing AI chatbots into everyday use. The two companies are now intertwined. Microsoft offers its vast resources, a need Scott outlined in an email, but OpenAI offers his cutting-edge AI expertise that Microsoft executives were hooked on. doing.
Scott said in an email that Google and OpenAI miscalculated exactly what they were trying to accomplish with their AI efforts. At the time, Google's startup DeepMind was trying to build an AI system that could play the Chinese board game Go, which Scott appears to have referenced.
“I was very negative about it,” Scott wrote. “That was a mistake.”
Scott marveled at how Google and OpenAI have built an entire infrastructure around advancing AI. In his email, he said he was surprised that Google and OpenAI were able to design the data center, source the silicon chips, and build the programming framework that makes all the work possible.
“When you take all the infrastructure that we built to build NLP; [natural language processing] “It made me take things more seriously because it was a model that couldn't be easily replicated,” Scott told Gates and Nadella.
Microsoft did not respond to a request for comment.
With the advent of AI, many of these resources are in high demand. AI requires extraordinary computing power to both run and train the models behind chatbots like Google's Gemini and OpenAI's ChatGPT. Data centers have become a popular technology investment and real estate asset, with some of the biggest investment firms vying to corner the market. There has been an unprecedented shortage of silicon chips, or semiconductors, as more and more companies try to buy up silicon chips in fear of running out. Its importance is best demonstrated by his Nvidia's legendary stock price surge over the past year. AI models also require vast amounts of cloud computing, and companies like Google and Microsoft have invested heavily in cloud computing over the past decade.
But when he sent his email in 2019, Scott recognized how important all these infrastructure upgrades would be to the development of the types of AI that are now commonplace. “We became very concerned trying to understand where the capability gaps were between Google and us when it came to training models,” Scott said.
Scott said at the time that it took Microsoft about six months to train one of its AI models because “our infrastructure wasn't up to the task.”
Microsoft also found itself lagging behind its competitors in terms of the talent it put into machine learning and AI research. Since then, AI-savvy employees have had no shortage of job offers (some with multi-million dollar salary packages) from companies eager to hire them. The hiring boom for AI talent will eventually extend beyond the technology sector to nearly every industry in the corporate world.
Scott said Microsoft had some “very smart” machine learning experts, but they didn't have the right tools to make a noticeable impact on deep learning, the complex training mechanism used to develop AI models. There was a lack of resources and personnel. This meant their work took longer than necessary, a worrying prospect in the looming AI arms race. “We are years behind in terms of competition. [machine learning] It’s huge,” Scott said.
Nadella, meanwhile, appeared to take Scott's broader concerns to heart. Nadella responded in a CC email to Chief Financial Officer Amy Hood that this is “why I want to do this.”