Microsoft’s Nadella Is Building an AI Empire

[00:00:01] Welcome to Tech News Briefing, it's Tuesday, June 18. I'm Zoe Thomas for The Wall Street Journal. Apps are constantly trying to engage users. One focused on learning languages is succeeding at hooking them and keeping them coming back.

[00:00:19] We'll find out how Duolingo taps into the power of streaks and well-timed nudges to win the battle for users' attention. And then, Microsoft CEO Satya Nadella bet the future of the company on the potential of artificial intelligence when he forged a groundbreaking partnership with OpenAI.

[00:00:38] But Nadella isn't content to rely on OpenAI to dominate in this new era. Our reporter Tom Dutton will join us to explain how the CEO is spreading his AI bets around and what that means for Microsoft.

[00:00:54] But first, messages from a cartoon owl are blowing up the phones of people trying to learn another language. Duolingo, the company behind the messages, is succeeding at getting millions of users hooked and keeping them engaged with its app.

[00:01:10] The company, which teaches over 40 different languages, is worth around $8.5 billion. Here to talk about how Duolingo became one of the biggest winners in the global battle for your attention is our columnist Ben Cohen, who writes The Science of Success.

[00:01:25] Ben, Duolingo users keep coming back every day to maintain their streaks. What makes it better than other apps at measuring and encouraging routines? The first is that they are relentless and rigorous about A-B testing everything. Even a few words can make a huge difference.

[00:01:44] And so oftentimes you will see them use the word habit and not goal. And that's because when they A-B test all of these alerts, they find that people are more willing to open alerts that show habits and not goals.

[00:01:57] But the other thing is that when you think about why we opened Duolingo, we are intrinsically motivated to do so, right? And when they're reminding us that we have invested 481 days in the app, you think, oh,

[00:02:10] I've already invested all this time for Duolingo because you have chosen to do this and you feel like you are getting better. You are improving or you just find it delightful. You choose to actively engage. How long are the streaks according to Duolingo?

[00:02:25] So Duolingo has about 30 million daily users. They say that about 20% of them or more than 5 million have streaks of at least a year. So 365 days, 70% of their daily active users have streaks of at least a week. And then there are some outliers who have streaks for many years.

[00:02:43] So the longest active streak on Duolingo right now is more than 4,000 days, which is more than 11 years. Okay, wow. Tell us about the algorithm that makes this possible. There are a lot of them. The big one for the notifications is what's called a multi-armed bandit algorithm.

[00:03:00] The goal of that algorithm is to send you the alert that you are most likely to open at any given time. So if you think about other companies and businesses that are sending you alerts, sometimes

[00:03:13] they're being sent to everyone around the world at the exact same time, even if you are not interested in it. That's not how Duolingo operates. Duolingo tailors alerts and personalizes them to maximize the chances that you are going to open it.

[00:03:28] And so the fact that they know exactly when you are using the app and how you are using the app, and then they can cater the experience of the app to you is really powerful.

[00:03:36] And it takes a lot of really fancy technology and AI and hidden algorithms that just you wouldn't think of a language learning app as being that sophisticated. That was our columnist, Ben Cohen. Coming up, Microsoft is spreading its AI bets around.

[00:03:52] We'll tell you why the tech giant isn't content simply relying on its partnership with open AI. After the break. AI may be the most important new computer technology ever, but AI needs a lot of processing speed and that gets expensive fast.

[00:04:12] Upgrade to the next generation of the cloud, Oracle Cloud Infrastructure or OCI. OCI is the single platform for your infrastructure, database, application development and AI needs. Do more and spend less like Uber, 8x8 and Databricks Mosaic.

[00:04:28] Take a free test drive of OCI at oracle.com slash wallstreet oracle.com slash wallstreet. Microsoft CEO Satya Nadella may be the ultimate company lifer. Over the last 32 years, he worked his way up through Microsoft's Bing search and cloud infrastructure businesses.

[00:04:51] Since taking over the top job in 2014, he has embraced other tech giants more than his predecessors, striking partnerships. He's also made big acquisitions. Now Nadella is building an AI empire. Our Microsoft reporter Tom Dutton joins us now with more. So Tom, what's Nadella's leadership style like?

[00:05:12] He's tried to run Microsoft more like a startup, which is a kind of difficult thing to do when you are running a company that is, you know, decades old and worth now trillions of dollars.

[00:05:23] But his idea when he took over as CEO in 2014 was to try to make the company more nimble, embrace more partners, take on new ideas, not be so resistant to things that originated outside the company.

[00:05:36] And the clearest way that that's really been achieved is in its current era with artificial intelligence. If you're an investor in Microsoft, you're pretty pleased with Satya. Share price has gone up tenfold since he took over as CEO.

[00:05:48] Microsoft made this groundbreaking partnership with OpenAI, the creator of ChatGPT. And just to note, News Corp, the owner of The Wall Street Journal, has a content licensing partnership with OpenAI. But Tom, how was the move received internally at Microsoft?

[00:06:03] The partnership with OpenAI has been in one sense transformational for the company. There's no way that you can look at what's gone on in terms of its share price increase, the cultural significance that the company has suddenly gotten.

[00:06:16] It's obviously it's products that it's put out built on top of OpenAI's technology, co-pilot the relaunch of Bing with this chatbot. Now they've got co-pilots for all their main software products like Word and Excel.

[00:06:29] The part of it that has been more difficult has been the integration of this outside technology. Microsoft had its own AI teams and they were building their own version of these large language models, which are at the center of this AI boom.

[00:06:44] And all of that stuff has been difficult inside Microsoft to kind of see this outside company being given all of the priorities, all of the resources in favor of what the company had built internally. Microsoft is building up its internal AI team too, though.

[00:06:59] How has the company been doing this? So that's been a more recent trend. Basically what happened is at the end of last year, OpenAI kind of imploded. If people remember that whole fiasco of Sam Altman, the CEO of OpenAI getting fired by

[00:07:12] the board, there was this weekend long tussle between the board and Sam Altman and Microsoft and Satya Nadella played a pretty major role in getting Sam reinstated as the CEO of OpenAI. As we've previously reported.

[00:07:27] But since then, there's been a shift at Microsoft, at least to a certain degree, over whether they should be completely reliant on OpenAI given that it can just be risky having an outside company running your most important initiative in AI.

[00:07:42] And earlier this year, Satya hired the CEO and a bunch of members of the team of this startup called Inflection, which is a competitor of OpenAI's. And they've integrated these people into the company under a new division called Microsoft AI.

[00:07:57] And one of the biggest initiatives that this division has is building up a new large language model that is designed to be at par with the capabilities of OpenAI's technology. And so overnight, it gave Microsoft a new infusion of talent and the possibility of a plan B.

[00:08:13] Tom, based on your reporting, what have you learned about how OpenAI CEO Sam Altman and Inflection co-founder Mustafa Suleiman interact in the development of AI products? Since Mustafa joined Microsoft and began heading up this division, he is one of the main points of contact between OpenAI and Microsoft.

[00:08:32] He and Sam have met multiple times since he joined, and that's according to our reporting and talking to people familiar with the situation. And Tom, what did the decision to invest so much in OpenAI mean for other departments in Microsoft?

[00:08:46] One of the biggest issues that came up when OpenAI was deemed the top priority at Microsoft was that they just get all the resources. Like the chips that are needed to train AI models, they are still somewhat in limited supply.

[00:09:01] They're definitely a couple of years ago were in very limited supply and Microsoft shifted almost all of their AI computing resources toward OpenAI. So if you were in a research division or any other AI team at the company, it has been really difficult.

[00:09:17] At the same time, the cost of all this stuff is extreme. The amount of new data centers that Microsoft has to put up in order to train AI models or to even run OpenAI's technology is high. And so they've had to make cuts in other areas.

[00:09:31] So it's been fairly distributed across the board at Microsoft as to what areas have been hit by this budgetary pinch. But hardware was one of the ones that I kind of was talking to people a lot about.

[00:09:43] There's just a lot of hardware projects that are given far fewer resources than they were before the whole AI push. It's not just OpenAI and inflection. Microsoft is making deals with multiple AI startups. What does it hope to gain from those? Well it's twofold.

[00:09:57] One is they run Azure, which is their cloud computing business, which is essentially a platform. And so they want to bring on as many different partners onto Azure as possible. They want to be competitive with Amazon, competitive with Google when it comes to the

[00:10:13] availability of all these different AI startups models on their platforms. But at the same time, it's also about making it clear to the world that Microsoft is not just the OpenAI business. They made this investment in this French AI startup called Mistral, and that's also being

[00:10:30] listed on Azure. They have a partnership with Cohere, which is another competitor with OpenAI that's listed on Azure. And in general, they are out there trying to redo the image of Azure, not just as this cloud computing company that's for old-timey, musty, enterprise giant companies, but also

[00:10:49] like a place that startups would want to go to. What have all these AI-focused changes meant for internal morale at Microsoft? It's been tough for a lot of people there. Bringing on that inflection team caused major disturbances in the group that was overseeing

[00:11:06] the Bing chat tool and the integration of that. And there were a number of top executives that we reported who left or are leaving because of this. And in general, you know, there's a lot of turmoil that happens when you take on a number

[00:11:17] of outside people and integrate them into very highly powered positions. There's not a small number of people that are just feeling just generally burnt out. That was our reporter Tom Dutton. And that's it for Tech News Briefing.

[00:11:29] Today's show was produced by Julie Chang with supervising producer Catherine Millsop. I'm Zoe Thomas for The Wall Street Journal. We'll be back this afternoon with TNB Tech Minute. Thanks for listening.