- Kevin Scott said Microsoft's partnership with OpenAI was a strategic "bet" on Sam Altman's company.
- The Microsoft CTO gave insight into its investment in the ChatGPT maker on the "Possible" podcast.
- Scott said Microsoft thought working with OpenAI would "push us to build better infrastructure."
Microsoft's chief technology officer said partnering with OpenAI was "basically a bet" on Sam Altman's company.
Kevin Scott gave insight into his decision-making that led to Microsoft's alliance with OpenAI in 2019 on an episode of Reid Hoffman's podcast "Possible" last week.
"This partnership with OpenAI was like basically a bet saying this particular team at the time also understood that this was a game of scaling, compute and doing incredible things with it in a very disciplined way," he said.
Scott, who's been Microsoft CTO since 2017, said he realized pretty early into his tenure that the progress of AI was accelerating.
For Microsoft to remain competitive, he knew it had to advance its infrastructure, which he started to think about how to build in 2018.
In an email sent to Microsoft cofounder Bill Gates and CEO Satya Nadella the following year, Scott said he was "very, very worried" about Google's AI capabilities at the time.
The email, which was recently made public as part of the Department of Justice's antitrust case against Google, had the subject line "Thoughts on OpenAI." It was sent just weeks before Microsoft announced its $1 billion investment in OpenAI and subsequent partnership. A large section of Scott's email was redacted.
Scott also told the podcast that the key drivers behind the partnership were knowing Microsoft couldn't do everything by itself, and realizing OpenAI could forecast what it could get from investing in computing.
"And I was like, 'if we work with them, they will push us to build better infrastructure, and we can enable them to do their best work'," he said.
Microsoft is building its own large language model (LLM) to take on those from Google and OpenAI, The Information reported on Monday. The new model, called MAI-1, will have roughly 500 billion parameters, the report said. OpenAI's GPT-4 is thought to have about one trillion parameters.
Scott appeared to confirm the report about MAI-1 in a LinkedIn post on Monday: "Microsoft, and the teams making and operating things on occasion need to do their own custom work, whether that's training a model from scratch, or fine tuning a model that someone else has built."
He added: "There will be more of this in the future too. Some of these models have names like Turing, and MAI. Some, like Phi for instance, we even open source."
Microsoft didn't immediately respond to a request for comment from Business Insider, sent outside normal working hours.
