Former Microsoft CTO Nathan Myhrvold on Bill Gates, AI buzz, and his 2,500-page pastry book

This week on the GeekWire Podcast, we’re featuring highlights from a live interview with Nathan Myhrvold, CEO of Intellectual Ventures and former chief technology officer at Microsoft.

Myhrvold worked at Microsoft from 1986 to 2000, where he laid the groundwork for Microsoft Research, recruited top computer scientists, and played a key role in shaping the company’s technology strategy.

Since leaving Microsoft, he has worked across fields including energy, science, physics, paleontology, photography, and high-tech cuisine.

In this conversation, recorded at Town Hall Seattle as part of GeekWire’s Microsoft@50 event, Myhrvold shares his thoughts on the rise of AI, his longtime collaboration with Bill Gates, the future of energy, the secrets of Microsoft’s success, and what’s next in his Modernist Cuisine book series.

Listen below, and continue reading for highlights from Myhrvold’s remarks, edited for context and clarity.

On predicting Microsoft’s rise: In 1987 I told [Bill Gates] that Microsoft would be the most valuable company on Earth, and he’d be the richest man in the world, and it would take 10 years. And I was totally wrong. It took three… I had not figured that Sam Walton would die.

On Bill Gates’ brutally honest feedback: I remember Bill said, ‘Well, you’re new here, so I’ll forgive the stupidity of your remark.’ That was the type of supportive feedback [that he gave].

On Microsoft’s willingness to admit mistakes: They would admit they were wrong. For some people, especially big-shot CEOs, that’s really hard. But for Bill, it was more embarrassing to not admit it.

On the boom-and-bust fashion cycles of AI: AI is like clothes — it goes through cycles of being in fashion and out of fashion… When speech recognition became successful, they called it ‘speech recognition’ — that used to be AI, until it worked.

On where AI stands today — and how far it still has to go: Today, [AI] is a lot like personal computers in the 1980s… good for a bunch of things… but its potential is enormously higher. And that will require a whole lot of work by a whole lot of folks.

On the ‘miracles’ still needed for human-level AI: AI doesn’t yet have the ability to create new abstract concepts and reason about them. That’s at least one miracle that needs to be figured out. I’ve variously thought there are three to five miracles. It could happen tomorrow — or maybe it already happened tonight, and they just haven’t told us.

Why he’s not losing sleep over AI doom scenarios: Humans love really scary, nasty villains that aren’t actually real. Sauron, the Night King — those weren’t really going to get us. AI overlords destroying us is very similar… it’s a story you can get excited about, but we all know there’s no AI overlord outside that’s going to get us.”

On energy use, AI, and the global demand for more power: “The average American uses about 12 kilowatts — it’s like you had 12 toasters running 24/7… The poor world wants to get rich, and the rich world wants to do more things that require power, like AI.

On using AI to analyze thousands of pastry recipes: I’m writing a big, 2,500-page book on pastry… And I use AI in that. I’ll say, ‘I think this assumption isn’t necessary,’ and ChatGPT will always say, ‘You’re absolutely right.’ Oh my God, it’s learned how to butter me up.

On the surprising reactions to his latest food project: I’m writing a big, 2,500-page book on pastry… Literally, last week, I said to someone, ‘I’m writing a big book on pastry,’ and they immediately said, ‘It must be hard writing on pastry.’ And I didn’t know if they were pulling my leg or what.

Subscribe to GeekWire in Apple Podcasts, Spotify, or wherever you listen.

Audio editing by Curt Milton.