But as C.E.O. of the resurgent Microsoft, he is firmly at the center of the A.I. revolution. We speak with him about the perils and blessings of A.I., Google vs. Bing, the Microsoft succession plan — and why his favorite use of ChatGPT is translating poetry.

You're the most wanted man on the planet.

Marvel Studios Secret Invasion, now streaming exclusively on Disney Plus.

Most CEOs of big technology firms are not loving every day right now.

They have been facing all sorts of headwinds and backlash.

But you can see why Satya Nadella might be the exception.

He has worked at Microsoft for more than 30 years, nearly 10 as CEO.

At the start of the personal computer era, Bill Gates' Microsoft was a behemoth, eager to win every competition and crush every rival.

But the internet era put the company on its heels.

Newer firms like Google, Facebook, and Amazon were more nimble, more innovative, and maybe hungrier.

Jeff Bezos of Amazon would reportedly refer to Microsoft as a country club.

But under Nadella, Microsoft has come roaring back.

He invested heavily in what turned out to be big growth areas like cloud computing.

Microsoft has always been in the business of acquiring other companies, more than 250 over its history.

But some of the biggest acquisitions have been Nadella's, LinkedIn, Nuance Communications, and if regulators allow, the gaming firm Activision Blizzard.

And there have been many more key acquisitions like GitHub, where computer programmers store and share their code.

Once again, Microsoft is a behemoth, the second most valuable company in the world, trailing only Apple.

Its stock price is up nearly 50 percent since the start of 2023.

But that's not even the reason why Microsoft has been all over the news lately.

They're in the news because of their very splashy push into artificial intelligence in the form of chat GPT, the next-level chatbot created by a firm called OpenAI.

Microsoft has invested $13 billion in OpenAI for a reported 49% stake in the company.

And they quickly integrated OpenAI's tech into many of their products, including the Microsoft search engine, Bing.

For years, Bing was thought of as something between footnote and joke, running a very distant second to Google.

But suddenly, Bing with chat GPT is on the move, And Google is trying to play catch up with its own chatbot called Bard.

So how exactly did Satya Nadella turn the country club into a bleeding-edge tech firm with a valuation of more than $2.5 trillion?

Our mission, Stephen, is to empower every person and every organization on the planet to achieve more.

And so as the world around us achieves more, we make money.

I mean, I assume you actually believe that.

You have to have a business model that is aligned with the world around you doing well.

Today on Freakonomics Radio, we speak with Satya Nadella about the blessings and perils of AI.

We talk about Google and Heidegger, about living with pain, and about Microsoft's succession plan.

No, it'll be nothing like that, Nadella promises.

This is Freakonomics Radio, the podcast that explores the hidden side of everything, with your host, Stephen Dubner.

I spoke with Satya Nadella 1 afternoon earlier this month.

I was in New York and he was in his office at Microsoft's headquarters near Seattle.

It's fantastic to have a conversation again.

We first interviewed Nadella in 2017 for a series called The Secret Life of a CEO.

Even then, he was extremely excited about AI.

At the time, Microsoft was high on a virtual reality headset called the HoloLens.

Think about it, your field of view, right?

What you see is a blend of the analog and digital.

The ability to blend analog and digital is what we describe as mixed reality.

There are times when it'll be fully immersive, that's called virtual reality.

Sometimes when you can see both the real world and the artificial world, that's what is augmented reality.

But to me, that's just a dial that you set.

I mean, just imagine if your hologram was right here interviewing me as opposed to just on the phone.

Back then, Nadella cautioned that there was still a lot of work to do.

Ultimately, I believe, in order to bring about some of these magical experiences and AI capability, we will have to break free of some of the limits we're hitting of physics, really.

The limits of physics haven't been broken yet, and the HoloLens has not been the hit that Microsoft was hoping for.

But Nadella's devotion to AI is paying off big time in the form of ChatGPT, which quickly captured the imagination of millions.

GPT stands for Generative Pre-trained Transformer, and CHAT-GPT is what is known as a large language model, or LLM.

It takes in vast amounts of data from all over the internet so it can learn how to read and answer questions very much like a human, but a really, really smart human or perhaps a million smart humans.

And the more we ask ChatGPT to answer questions or summarize arguments or plan itineraries, the more finely tuned it gets, which proves at the very least that we humans are still good for something.

And what's the relationship between ChatGPT and Bing?

Basically, Bing is part of ChatGPT and Chat is part of Bing.

So in either way, it doesn't matter which entry point you come to, you will have Bing.

So Satya, I asked ChatGPT for some help in this interview.

I said, I'm a journalist interviewing Satya Nadella and I want to get candid and forthright answers.

You know, I just didn't want corporate boilerplate.

And what chat told me was to do my homework, which, you know, I did.

I usually do that to ask open-ended questions, which I typically try to do.

But 1 that hung me up a little bit was I need to build rapport.

Now we have a relatively short time together today.

Are there any shortcuts to building rapport?

I knew that you're a big cricketer, you played as a kid, I knew you cared more about cricket than schoolwork as a kid, but no, I blew it.

That's too bad, because there's a world chess championship starting tomorrow.

I was going to ask you about it, but hey, look, Your love for economics builds me an instant rapport.

I'd like you to walk us through Microsoft's decision to bet big on OpenAI, the firm behind ChatGPT.

There was an early investment of a billion dollars, but then much, much more since then.

I've read that you were pretty upset when the Microsoft research team came to you with their findings about OpenAI's LLM large language model.

They said that they were blown away at how good it was and that it had surpassed Microsoft's internal AI research project with a much smaller research team in much less time.

Tell me if what I've read, first of all, is true.

Were you surprised and upset with your internal AI development?

Yeah, I think that this was all very recent.

This was after GPT-4 was very much there.

And then that was just mostly me pushing some of our teams as to, hey, what did we miss?

You know, there were a lot of people at Microsoft who got it and did a great job of, for example, betting on OpenAI and partnering with OpenAI.

And to me, 4 years ago, that was the idea.

And then as we went down that journey, I started saying, okay, let's apply these models for product building, right?

The first real product effort which we started was GitHub Copilot.

And, you know, quite frankly, the first attempts on GitHub Copilot were hard because, you know, the model was not that capable.

But it is only once we got to GPD 3, when it started to learn to code, that we said, oh wow, this emergent phenomena, the scaling effects of these transformer models are really showing promise.

Nadella may be underplaying the tension between Microsoft and OpenAI, at least according to a recent Wall Street Journal article called the Awkward Partnership Leading the AI Boom.

It describes, quote, conflict and confusion behind the scenes.

And because the OpenAI deal is a partnership and not an acquisition, the journal piece makes the argument that Microsoft has influence without control as open AI is allowed to partner with Microsoft rivals.

Still, you get the sense that Nadella is excited about the competitive momentum ChatGPT has given Microsoft, as you can tell from this next part of our conversation.

Google still handles about 90% of online global search activity.

An AI search-enabled model is a different kind of search, plainly, than what Google's been doing.

How do you see market share in search playing out via Bing, via ChatGPT in the next 5 and 10 years?

And I'm curious to know how significant that might be to the Microsoft business plan overall.

This is a very general purpose technology, right?

So beyond the specific use cases of Bing Chat or ChatGPT, what we have are reasoning engines that'll be part of every product.

In our case, they're part of Bing and ChatGPT, they're part of Microsoft 365, they're part of Dynamics 365.

And so in that context, I'm very excited about what it means for search.

After all, Google, as you said rightfully, they're dominant in search by a country mile.

We've been at it to sort of say, hey, look, our time will come where there will be a real inflection point in how search will change.

We welcome Bing versus Bard as competition.

It'll be like anything else, you know, which is so dominant in terms of share and also so dominant in terms of user habit, right?

We also know that defaults matter and obviously Google controls the default on Android, default on iOS, default on Chrome.

And so they have a great structural position.

But at the same time, whenever there is a change in the game, it is all up for grabs again, to some degree.

And I know it'll come down to users and user choice.

If We finally have a competitive angle here and so we're going to push it super hard.

What are some of your favorite uses, personal or professional, for ChatGPT?

The thing that I've talked about, which I love, is the cross-lingual understanding.

You can go from Hindi to English or English to Arabic or what have you, and they've done a good job.

If you take any poetry in any 1 language and translate it into another language, In fact, if you even do multiple languages, so my favorite query was, I said I always as a kid growing up in Hyderabad, India, said, I wanna read Rumi translated into Urdu and translated into English.

And 1 shot it does it, But the most interesting thing about that is it captures the depth of poetry.

So it finds somehow in that latent space meaning that's beyond just the words and their translation.

You, the CEO of a big tech firm, is saying that 1 of the highest callings of chat DPD or large language model is the translation of poetry.

I mean, I know you love poetry, but what excites you more about that than more typical business, societal, political, economic applications?

I remember my father trying to read Heidegger in his 40s and struggling with it.

And I've attempted it a thousand times and failed.

And he's written this essay somebody pointed me to.

Somebody said, oh, you got to read that because after all, there's a lot of talk about AI and what it means to humanity.

But I must say, you know, going and asking Chad GPT or Bing Chad to summarize Heidegger is the best way to read Heidegger.

According to Chad GPT, Heidegger himself would not have been a fan of AI.

In Heidegger's view, Chatt tells us, technology, including AI, can contribute to what he called the forgetting of being.

After all, philosophy and poetry will likely not be the main use cases for AI.

So, after the break, we talk about potential downsides of an AI revolution and the degree to which Microsoft cares.

I want all 200, 000 people at Microsoft working on products to think of AI safety.

Last month, a group of leaders from across the tech industry issued a terse one-sentence warning.

Mitigating the risk of extinction from AI should be a global priority alongside other societal scale risks such as pandemics and nuclear war.

The extinction they are talking about is human extinction.

Among the signatories were Sam Altman, the CEO of OpenAI, and 2 senior Microsoft executives.

Altman, Satya Nadella, and other executives from firms working on AI recently met with President Biden to talk about how the new technology should be regulated.

I asked Nadella where he stands on that issue.

I think the fact that we are having the conversation simultaneously about both the potential good that can come from this technology in terms of economic growth that is more equitable and what have you.

And at the same time that we're having the conversation on all the risks, both here and now and the future risks, I think it's a super healthy thing, right?

Somebody gave me this analogy, which I love, right?

Just imagine when the steam engine first came out, if we had a conversation both about all the things that the steam engine can do for the world and the industrial production and the industrial revolution and how it'll change livelihoods.

And at the same time, we were talking about pollution and factory filth and child labor, we would have avoided more of a 100 years plus of terrible history.

So then it's best to be grounded on what does the risk framework look like, right?

If AI is used to create more disinformation, that's a problem for our democracy and democratic institutions.

Second, if AI is being used to create cyber attacks or bioterrorism attacks, that's a risk.

If there is real world harms around bias, that's a risk.

Or employment displacement, that's a risk.

In fact, those were the 4 even the White House was upfront on and saying, hey, look, how do we really then have real answers to all these 4 risks?

So in terms of, for example, take disinformation.

Can we have techniques around watermarking that help verify where did the content come from?

When it comes to cyber, what can we do to ensure that there is some regime around how these frontier models are being developed.

Microsoft itself has been working on provisions to best govern AI.

For instance, safety brakes for AI systems that control infrastructure, like electricity or transportation, also a certain level of transparency so that academic researchers can study AI systems.

What about the doomsday scenario wherein an AI system gets beyond the control of its human inventors?

Essentially, the biggest unsolved problem is how do you ensure both at sort of a scientific understanding level and then the practical engineering level that you can make sure that the AI never goes out of control.

And that's where I think there needs to be a CERN like project where both the academics along with corporations and governments all come together to perhaps solve that alignment problem and accelerate the solution to the alignment problem.

But even a CERN-like project after the fact, once it's been made available to the world, especially without watermarks and so on, does it seem a little backwards?

Do you ever think that your excitement over the technology led you and others to release it publicly too early?

No, I actually think, first of all, we're in very early days and there has been a lot of work.

See, there's no way you can do all of this just as a research project.

In fact, if anything, that, for example, all the work we did in launching Bing Chat and the lessons learned in launching Bing Chat is now all available as a safety service, which by the way, can be used with any open source model.

So that's, I think, how the industry and the ecosystem gets better at AI safety.

But at any point in time, anyone who's a responsible actor does need to think about everything that they can do for safety.

In fact, my sort of mantra internally is the best feature of AI is AI safety.

I did read, though, Satya, that as part of a broader, a much broader layoff earlier this year that Microsoft laid off its entire ethics and society team, which presumably would help build these various guardrails for AI.

From the outside, that doesn't look good.

At the same time, I saw all the headcount that was increasing at Microsoft.

It's kind of like saying, hey, should we have a test organization that is somewhere on the side.

I think the point is that work that AI safety teams are doing are now become so mainstream, critical part of all product making, that we have actually, if anything, doubled down on it.

So I'm sure there was some amount of reorganization and any reorganization nowadays seems to get written about and that's fantastic, we love that.

But to me, AI safety is like saying performance or quality of any software project.

1 particular concern about the future of AI is how intensely concentrated the technology is within the walls of a relatively few firms and institutions.

The economists, Daron Acemoglu and Simon Johnson recently published a book on this theme called Power and Progress, our 1000-year struggle over technology and prosperity.

And here's what they wrote in a recent New York Times op-ed.

Tech giants Microsoft and Alphabet Google have seized a large lead in shaping our potentially AI-dominated future.

History has shown us that when the distribution of information is left in the hands of a few, the result is political and economic oppression.

Without intervention, this history will repeat itself.

Their piece was called Big tech is bad, big AI will be worse.

You could argue we are fortunate to have a CEO as measured as Satya Nadella leading the way at Microsoft, but of course he won't be there forever.

After the break, what does a Microsoft succession look like?

I'm Stephen Dubner, this is Freakonomics Radio.

Sathya Nadella grew up in India where his father was a Marxist civil servant.

Satya really did want to be a professional cricketer, and he wasn't a great student, but he did go on to study electrical engineering at university, then came to the States for a master's degree in computer science.

He worked for a couple of years at Sun Microsystems, then joined Microsoft in 1992.

He also got an MBA from the University of Chicago.

In his early years at Microsoft, he worked on the operating system Windows NT.

Windows products were how Microsoft made its big money, and the majority of the world's desktop computers still run on Windows.

But Microsoft famously missed the shift to mobile computing, and from 2000 to 2013, under CEO Steve Ballmer, they saw their stock price fall by more than

Satya Nadella inherited a lot of problems.

Officially, He is a professor of management studies at Yale and he runs the Yale Chief Executive Leadership Institute.

Unofficially, he is known as 1 of the world's leading authorities on CEOs.

Last year, he published a list of the best American CEOs.

Rather than try to come up with long lists of ways of vilifying predecessors, what Nadella did is he was able to be on a frontier at this exact same moment as the early investors in open AI as well as in reinventing their own artificial intelligence opportunities so that being surprised all might soar past everybody.

He got people excited about building a new future, investing 25 billion in R&D each year.

That's perhaps twice as much as the average pharma company invests, and that's amazing for an IT company to do that.

A big part of Nadella's success came from expanding Microsoft's footprint in cloud computing with their Azure platform.

Their footprint across the board at enterprise software was flourishing where he knew how to invest in Azure and a commercial cloud business where His revenues grew 42% over the past year.

I asked Nadella himself if he had been surprised by how valuable cloud computing has become for Microsoft.

Both surprised and not surprised in the following sense.

We were leaders in client-server, But while we were leaders in client server, you know, Oracle did well, IBM did well.

And so in fact, it shaped even my thinking of how the cloud may sort of emerge, which is that it'll have a similar structure.

There will be at least 2 to 3 players who will be at scale, and there will still be many other smaller niche players, perhaps.

So in that sense, it is not that surprising.

What has been surprising is how big and expansive the market is, right?

You know, let's think about it like, yeah, we sold a few servers in India, but oh my God, did I think that cloud computing in India would be this big?

No, the market is much bigger than I originally thought.

I have a fairly long and pretentious question to ask you.

There are economists and philosophers and psychologists who argue that most of us still operate under a scarcity mindset that might have been appropriate on the savanna a million years ago, but now we live in an era of abundance.

So, you know, rather than competing for scarce resources, we should collaborate more to grow the overall resource pool.

From what I know about your time as CEO at Microsoft, it seems you have embraced the collaborative model over the competitive model.

1 example being how nicely Microsoft now plays with Apple devices, whereas the previous administration didn't even want Microsoft employees owning Apple devices.

See all Freakonomics Radio transcripts on Applepodcasts

547. Satya Nadella’s Intelligence Is Not Artificial