Microsoft’s First CTO Says AI Is ‘Three to Five Miracles’ Away From Human-Level Intelligence

Nathan Myhrvold had a front-row seat to the rise of modern computing that few can match. As Microsoft’s first CTO, he helped shape its technology strategy during its formative years from 1986 to 2000, working closely with Bill Gates and leading groundbreaking research efforts. Decades after this departure, he reflected on his career and shared insights into how he views the state of artificial intelligence (AI)—spoiler: He’s impressed but feels additional “miracles” are needed to make it truly transformative.

“We live in a technological world,” Myhrvold said on Thursday at GeekWire’s Microsoft@50 event in response to how he views Microsoft’s significance in modern business history. “We all interact with technology constantly, and Microsoft was an absolutely fundamental, foundational part of that.” He explained that the company was initially ridiculed for its motto,” a computer in every desk and every home.” Today, that’s a reality but at a grander scale, and Myhrvold argues the “world’s a better place for having been empowered by that.”

However, to achieve that, “you needed there to be standards in software that could allow the hardware people to innovate as much as they possibly can and software to innovate separately from that, and yet let things be compatible. That’s what Microsoft ultimately was about.”

AI Is Like the PC of the 1980s

There’s no question that the technology landscape has rapidly evolved in the quarter century since Myhrvold worked at Microsoft. Still, he remains attuned to changes through his firm, Innovation Ventures, which invests in technology and energy development. So, what does he think about AI and where it’s headed? Put simply, don’t consider Myhrvold a doomsayer.

“There’s a persistent thing with the tech industry that people overestimate the short term and underestimate the long term,” he claims, pointing out that in the short term, companies and pundits tend to overhype the benefits their AI solutions can have. Still, he believes AI’s value lies in the long term. “I think AI has tremendous potential, and I would put AI today a lot like personal computers in the 1980s, meaning it’s good for a bunch of things. I’m not saying it’s bad, but I think its potential is enormously higher, and that will require a whole lot of work by a whole lot of folks.”

Myhrvold says he’s “entirely heartened” by AI’s prospect, dismissing naysayers—those who perpetuate the idea that progress will destroy us—as equivalent to Luddites. “What’s amazing is that fairly simple techniques and tremendous amounts of computing power actually achieved something that is able to mimic most parts of human language. That’s almost unbelievable,” he points out, contending that in addition to “Godfather of AI,” Geoffrey Hinton, being awarded the Nobel Prize in physics last year for his work on machine learning, an additional award should have been bestowed on this idea.

AI is Good But Still Needs More Miracles

Despite AI’s disruptive impact, Myhrvold argues that it still has a long way to go before it can truly match human capabilities. He explained why replicating the richness of human language is particularly challenging for AI:

“None of us here learn to read or learn to do anything by reading every book on Earth. We couldn’t possibly, and that’s largely because we have this very facile ability to generate new abstractions on the fly. If I call this thing a widget, and I said nothing about the widget, we can talk about widgets and you’ll know when referring to that, and we can reason about it. And that facility to create a new abstract concept, imbue it with meaning and properties, and then reason about it, AI doesn’t have that gap. That’s why we have to train it by having every example you can possibly imagine of different combinations of words. But I think we’ll get there—that’s at least one miracle that needs to be figured out. And I variously have thought there was like three to five miracles that need to be done.”

Microsoft’s former CTO didn’t elaborate on those so-called “miracles” and expressed uncertainty about when they could occur. They could happen “tomorrow, or maybe it’ll happen tonight, and they just haven’t told us, or it could…take another ten years.” Regardless, Myhrvold stressed what’s more important is how AI is being used—it’s “the work of harnessing that power to do something for you, and that’s what application programming is about.”

And while companies are embracing an AI-first mentality and launching agent chatbots to tackle the jobs to be done, Myhrvold doesn’t think we’ve seen the real AI use case developed yet. “We’ve only scratched the surface,” he declared.

AI Fortune Favors the Bold

When pressed to elaborate on these additional miracles needed to improve AI, he remarked that it’s difficult to ascertain when they would occur. Myhrvold argues that looking back, mostly no one would have guessed where AI has progressed to today, aside from a “small set of people at OpenAI and at Microsoft who believed that crazy proposition which is, if only we poured billions of dollars of computing at it, we’ll get there.” It’s also not as if OpenAI and Microsoft were the only ones working on this problem—many people have tried since the 1960s, but none found success.

Myhrvold credits Microsoft’s calculated risk for helping put the company in “one of the best positions in AI.” The company’s partnership with OpenAI is part of the learning process, recognizing that Microsoft won’t always have the best ideas. He believes the deal to invest billions of dollars to a company with four employees at the time was another calculated risk and one that made sense: “They bet on the right one.”

He shrugs off concerns about AI threatening human existence, chastising doomsayers. “Here’s the other idea that humans love: We love having really scary, nasty villains that aren’t actually real. So Sauron (from “The Lord of the Rings”), none of us felt personally threatened by Sauron. The Night King and his army of the undead weren’t going to get us either. And so, when people conjure up these stories about AI overlords destroying all of us, it’s very similar to that. It’s a story you can get very excited about. You can say you’re getting frightened about [it], but really, we all know we’re going to go home, and there’s no AI overlord outside that’s going to get us.”

Myhrvold’s perspective offers a mix of optimism and pragmatism. While he acknowledges the challenges ahead, he sees AI’s trajectory much like the early days of personal computing. His message is that progress won’t come from fear but from bold investments, experimentation, and a willingness to embrace the unknown. Those willing to take calculated risks—like Myhrvold thinks Microsoft did—may be the ones shaping the future of AI.

Featured Image: Former Microsoft CTO and Intellectual Ventures co-founder Nathan Myhrvold gestures on stage at GeekWire's Microsoft@50 event on March 20, 2025. Photo credit: Ken Yeung