Prominent Silicon Valley VC engages in a conversation with Nadella Latest insights on AI productivity and agency.

Noteworthy Silicon Valley Venture Capitalist Engages in Candid Discussion with Nadella on Latest AI Insights for Boosting Productivity and Empowerment.

Source: Newin

This week, Microsoft Chairman & CEO Satya Nadella not only appeared at the first OpenAI Developer Conference, but also participated in a FireChat in San Francisco with renowned Silicon Valley investor Elad Gil to discuss the productivity revolution, code, and AI agents brought about by the AI wave. They also talked about how Satya himself led Microsoft to become a $2.5 trillion tech giant.

• everLianGuaiy Real-time Payment Layer 2 based on Arweave and SCP
• Cryptocurrency Master McAfee’s Widow Unaware of Husband’s Cause of Death, Relies on Gig Work for Survival.
• Gaming Giant WeMade’s Infamous Token ‘Relisted’ in South Korea Drama Reloaded!

Here’s the full transcript of their conversation, enjoy~

Elad Gil:

When you joined Microsoft in 1992 and became CEO in 2014, it became different under your leadership and became one of the most exciting and interesting companies in the tech industry, thanks to major initiatives like LinkedIn and OpenAI Games. First of all, we would love to hear what Microsoft was like when you took over and how you thought about the culture, direction, strategy, and how you influenced it over time?

Satya Nadella:

First of all, it’s great to be here with all of you. The fact is, I have now spent over thirty years at the same company, almost my entire career. And now I have better language to describe it. But I think companies, and I’ve been a founder of a number of them, do go through these moments of redefinition. And I learned this idea from Reid Hoffman, which I think is a great frame. And it’s not talked about enough.

When I became CEO in 2014, as a long time employee of the company, I felt that we needed to redefine the company, go back to the essence of our mission and our purpose, and then pursue new, bold things on top of that mission. In fact, when I joined the company in 1992, we used to talk about the mission of “a personal computer on every desk and in every home.”

In retrospect, that seems easy, you can put it on a spreadsheet and calculate it, and so on. But by the late 90s, at least in the developed world, we had pretty much achieved that goal. And from then on, I felt that we were a little bit like, “Okay, now what? Do we return all the money to the shareholders? So I went back to the origins of the company. After all, we were a tool company to start with.

Bill Gates built the basic interpreter, and by 2023, the need for tools and platforms would be even greater, that’s our core, building technology so that others can build more technology, let’s get back to that mission. That’s been very helpful; the other aspect is culture, which you touched on.

I vividly remember Microsoft in the late 90s, walking through the campus, and everybody’s saying, “Wow, we must be a gift to humanity because we are so good now, and the market also recognizes that. But clearly, in some sense, that was a beginning of an end, where arrogance starts to gain over humility, rather than standing on the things that got you successful in the first place, which is customer feedback loops, a thirst for learning, trying, doing things that clearly, startups do.

I love learning about this cultural framework from Cattle Duak at Stanford University, which is called growth mindset. I mean, God, we don’t want to be those know-it-all people, we want to be the ones who are willing to learn. It’s really a great framework because, first of all, it’s not some new doctrine proposed by a new CEO. I think it’s something that is inherent in people as part of being human, and it helps us become better parents, better partners, better colleagues, better neighbors, better leaders.

So, somewhere in between those two points, the mission of the company was redefined. And let me tell you, especially for you guys, I don’t think that founders instinctively have this ability, because they created something out of nothing. And then there’s the personality cult, which gets passed down. But at some point, if your company is going to transcend you, the mission has to be much bigger than the culture and the personality. So I think that CEOs, especially exemplary CEOs like myself, should focus more on the mission. Of course, it also gives you the power to make the right decisions. But you have to make the right decisions in terms of strategy, choice, and execution, because, you know, at least it gives you a better chance of achieving your goals.

Elad Gil:

How do you view the ability to pursue strategy in today’s environment? I know M&A (mergers and acquisitions) may be a bit more challenging for various reasons. AI and other technologies are making waves. What advice do you have for founders or people running successful businesses in thinking about their direction and thinking about tools like M&A or organic growth?

Satya Nadella:

I think whether it’s a startup or a big company, ultimately we are all fundamentally betting on organic growth, right? Even at Microsoft, I think most of the content that gets written is our non-organic growth, but when I think about most of the breakthroughs and the biggest revenue streams, frankly, they’re mostly partnerships.

Organic partnerships are another thing that people don’t talk enough about. I think it starts with a great partnership. In fact, I grew up in the Gates-Grove model, which I love, right? Intel and Microsoft creating the personal computer ecosystem. It’s a wonderful partnership. Equally important, in a partnership, it is key to make sure that if one partner becomes too greedy, then it becomes unstable. But if you can cultivate great partnerships, that’s another fantastic source of growth for the company.

Yes, non-organic partnerships, and then strategic M&A, all three are important for the company. Yes, I think we are going through some, I would say, global regulatory adjustments now, is the best way to describe it, to understand whether we should allow big companies to just do M&A because they’re big, and I think that’s going to create a chilling effect. And that’s not good for the creation and dynamism of startups.

Let’s assume this problem is solved, and I think we will always look at these three things: First, focus on creating an excellent organic plan, because that is something you can always control; partnerships are something I would say don’t treat it like a press release, but I think it can definitely be a long-term stable partnership, where you win, I mean a triple win, the customer wins, the partner wins, you win.

In fact, the best partnerships are when you start caring about your partners, making sure they get economic surplus, because if everything is about a zero-sum game, you know, to some extent sometimes in Silicon Valley culture, we have a bit of an overemphasis on zero-sum games, but in reality, there are very few zero-sum games, when you think about it, but we are very obsessed with everything being zero-sum. That’s where I think a little subtle, nuanced difference can be helpful.

Elad Gil:

A major transformation that is currently happening is the role Microsoft is playing in it, and that is the transformation of AI. These Transformer-based models, diffusion-based models, have truly changed the trajectory of what we can do with these machine systems. You’ve been at the center of some of the biggest technology waves in the past 30 years, including the internet, cloud, PC revolution, and mobile, among other things. How important do you think AI is relative to all these other trends?

Satya Nadella:

I think it is as important in some ways as any one of the things you mentioned. You know, I’m trying to get back to what exactly is happening and how we should relate to it in terms of any business or product development. I think there are two major changes, right? One of them is, I think we’re going to think about application interfaces in a completely different way, right? We have more tools, I think we can rethink this, right? Chat started as chat, but it quickly escalated beyond that, it’s multimodal, and for us, applying it to some extent would be very interesting.

So in a way, it reminds me of at least what happened in our history, when I joined Microsoft in 1992, we had just released Word, Excel, Powerpoint, or Publisher, and there was a new application every day because Windows 3 had just started. It felt like back then, you could rethink existing categories, use new interfaces, but we also knew that the new platform was not just a redesign of the old platform.

But there is also the question of what can this interface create for businesses that don’t exist yet? In fact, that is what interests me the most. Obviously, mobile devices have a lot of functionality on desktops, but they also created companies like AI rbnb and Uber. I think in terms of interfaces, we should think about what AI-first interface applications are, and I have started using them. I don’t think we have fully mastered it yet, but I think we are getting close.

I think the second one is also more about other technologies that people don’t talk about, I don’t know why, when we talk about our big paradigm, it’s relational databases. This, I mean, what is it? It’s the least. I think in the field of digital technology, all other fields are about the process of digitizing people, places, and things, right? This is another 70 years. Every day we wake up, more places, people, and things are being digitized.

Relational algebra and relational databases help us reason about them in an interesting way. I think we now have a new pattern recognizer or a new inference engine, which is accomplished through neural algebra.

I think these two factors, Elad, if you take an existing category or have a new way of thinking about the user interface, if you have a new way of thinking about all the data and knowledge in the world, making cross-connections, so to speak. If you think about applications in this way and build a new world-focused business, then I think it’s at least as important as any of the things you mentioned.

Elad Gil:

A lot of people are now talking about trends that, at least in my opinion, may be too early, but could be very interesting in the future, and these trends cover both aspects you mentioned, namely the shift towards an Agent-driven world. People talk about ultimately having Agents that represent us, these Agents will interact on our behalf with applications, companies, governments, etc. How much trust do you have in this Agent-driven action or Agent-driven interface future world? If such a world does exist, when do you think it will arrive?

Satya Nadella:

I believe in it very much. In fact, at least in our world, one of the design patterns that I really like, that we chose to build for ourselves and propagate as a pattern for anyone else, is this Copilot.

First, there is a person in the loop. Human agency is important. So, the first example is something like Github Copilot, which is like a copilot embedded in the canvas of your application. They work together to help you with your tasks, like completing your coding or anything else. Then we have now propagated it into knowledge work with Microsoft 365 Copilot across all of our platforms. But ultimately, when I think about what we’ve done with Bing Chat, I think Bing Chat is basically a web copilot, and this M365 Copilot is a work copilot.

So a web copilot and a work copilot might be some kind of a general-purpose Agent. At least that’s how I envision it. But it needs an important capability, which is the ability to converse with other Agents, like customer service agents, travel agents, to get things done. Some of that might be interruption-driven, meaning it might be autonomous, but it will also come back to you for decisions. So, I think one of the key runtimes of our time will be a multi-agent runtime.

We have something called Autogen in open source, and it’s gotten some nice attention. So, we’re building something similar beneath our Copilot. Obviously, today OpenAI released some very interesting things, including GPT, which is an early-stage Agent on top of Chat GPT itself. They even have an Agent API.

These will all be put into our Copilot ecosystem. So I think, yes, for your basic premise, people, I think this idea, people will have agents, these agents will collaborate with each other. There will be some kind of super app, whoever can crack it, there will be several runtimes, people naturally tend to multi-agent frameworks.

Elad Gil:

You also mentioned that you are very excited about the business potential of these things, what kind of transformation will they bring. Obviously, Microsoft has done some very innovative things through Github Copilot and other areas you mentioned. Is there a business world area that Microsoft is not directly involved in that you think is most likely to undergo AI transformation in the short term?

Satya Nadella:

That’s an interesting way to phrase it. If I knew, I might already be doing it. But maybe a little deeper. You see, one point I insist on is that I think the technology industry has reached a point where we should directly translate our celebrated technological advances into broad economic growth. Because I believe there are some very real criticisms, some of which I think are valid, such as: you talk a lot about these technologies.

Where is the economic growth happening? I mean, the last time we checked in the developed world, after adjusting for inflation, we may be at zero or negative growth. So economic growth is really needed, while also preserving the health of the planet and a more equitable economic growth. In terms of growth itself, there are many other things that I think are also our core responsibility.

But having said that, we must drive economic growth, and that’s where I get excited about Github Copilot, when we can make things like software development more joyful, what we have already done, I mean, like what I just copied and pasted from various places, being distracted, rather than that, let’s stay focused and see how you can bring productivity back to software developers.

But what’s interesting is that the more we study the impact of Github Copilot on organizations, the more we find that if you make software development faster, the workflows of all the other functions around you will change. In fact, when a salesperson makes a pull request, that’s my thing, for example, wow, this is a different organization. So now you can imagine an entire organization running at different speeds. So I think we are on the forefront of discovering what this means, both in terms of these productivity enhancers within each function and in terms of what the workflow is if you have it.

Like you said, agents, I think we have to discover. If you remember the mid-90s, when we first started placing things like system records, such as CRM and ERP, we were also talking about something called business process reengineering. It’s also about a new way or doctrine, about how to run a business, like you wouldn’t have five finance departments, or you wouldn’t have costs in manufacturing that are not captured in financial accounting. So some business practices will have to change. So that’s why I think it might be interesting how people perceive a vertical industry or business process can be very different. I think a lot of people think, can there be a large enterprise that focuses only on a vertical industry or a business process? I think there can be. It all depends on how much economic surplus you can create with this technology.

Elad Gil:

You previously mentioned the importance of partnerships and how they have become a key part of Microsoft’s ongoing strategy, collaborating with different partners over time. One thing that I find really exciting is how you have developed a deep partnership with OpenAI, their closed-source models, and the work around the open-sourcing of llama. How do you view Microsoft’s interaction with both open source and closed source, and how do you see AI evolving over time, especially in relation to these partnerships?

Satya Nadella:

I mean, it’s not an either-or situation for us. What I mean is that GitHub exists primarily because there is support for the open-source ecosystem. So it’s not something we take lightly. Therefore, as a company, we will always be supportive of it.

As it turns out, I think most people are not aware that we are one of the biggest contributors to open source. Today, we may be the biggest contributor to Linux. So, now it’s ingrained in us fundamentally. But that doesn’t mean we don’t have a bunch of proprietary closed systems and revenue streams. So for us, it’s almost part of our business.

Then the question is, what’s the best way to meet the needs of the developers, companies, and organizations we serve? So that’s why even though OpenAI is our primary partner at the cutting edge of models, we also have a lot of open-source models, including our own, that we are very excited about and that we contribute to, just like I love these SLMs small language models and PSI. So we will always be making contributions and supporting, ensuring that developers have choices, and you will also see a mix of usage in our own products.

Elad Gil:

I guess in the emerging criticism of open-source models and advanced large language models, there is a concern about security, and I think people often confuse three types of security. The first one is what I almost consider text safety, which involves misinformation, bias, hate speech, and so on. The second is physical safety, whether AI will be used to derail trains or create viruses and such. And then there is existential safety. Are you worried about some sort of conflict in the future with some form of AI species or agent, vying for resources or other possibilities? How do you view these criticisms and how do you view AI safety evolving over time?

Satya Nadella:

I like this categorization, in fact, I remember a few months ago when we spoke, you brought up this categorization and I really liked it. I think one thing that we need is to find a solid way to talk about these three levels. Because when we talk about safety, it may mean we are talking about an existential issue or deepfakes in elections, and so on.

I think let’s analyze them in this way. My sense is, first and foremost, what we really need to focus on are any real-world harms caused by the deployment of AI, because those problems have not yet been solved. In fact, especially in democratic countries, what concerns me the most is elections and whether our democratic process will be influenced by the use of some AI. I think if there are issues, then we should rightly be held accountable.

After all, that’s the criticism of what’s happening on social media. Now everyone says we won’t let that kind of situation happen again, but the question is how do we do it? What should the government do? What should civil society do? What should companies do? Frankly, I think we have to realize in some way that this is ultimately a social choice. It’s not something any one company can make all the decisions on. It’s social norms and decisions. So this is actually the most complex process we have to be involved in.

On the other hand, the two other levels, existential risk and physical safety, have time to address. Secondly, I think there are more engineering solutions feasible in this area. I mean, if you have a program of self-improvement where we’re losing control, for example, last I checked, people in other engineering fields knew about control theory and how to think about it. So we need to be able to learn something from other fields and apply it. But I’m glad the conversation is happening about existential risk.

Meanwhile, the mid-level issues, which you mentioned, I also like, namely the question of deploying AI in the real world, for example, if it’s going to actually act in the world, then that may be a question that needs to be thought through more carefully. In this case, a risk-based regulatory framework is also a helpful tool. Just like cars are deployed, but they have regulations different from many other regulations. So, in areas like healthcare, financial services, and car safety, where there are already regulations, we can incorporate AI into the same regulations.

So I think there may need to be more risk-based approaches on the mid-level, more research funding, frankly, for the harms of the present world, whether it’s bias or election interference, we need to think about what social norms are.

This is very important for us as a democratic society, especially in the United States, and I hope that we can take on this responsibility, I really hope there will be an election, a politician who can win an election with a clear vision. That’s where I feel uneasy when tech and tech leaders speak up about it. I didn’t elect tech leaders to represent me as an elected official. So, I hope more elected representatives can win elections by presenting a genuine vision and defining the norms in which we participate.

Elad Gil:

That makes a lot of sense. I guess related to that is the potential for AI to contribute to global fairness. When I look at the areas it can impact, one of them is healthcare. You can see the tremendous performance of Google’s Med LianGuailm 2 model in the healthcare field and its ability to generate good results based on health-related criteria.

Then there’s education too, with Microsoft having Minecraft, which is one of the most popular tools I’ve seen parents engage with. When I think about the ability to impact education, I think about going back to something that can customize or interact with your child in a rich way, an Agent-driven world that can provide guidance and education, and I’m interested in your thoughts on these areas.

Satya Nadella:

I believe that education, health, and finance are three important areas. Of course, we should strive for them. First, in terms of education, last week in the United Arab Emirates, I saw that their health department has an AI minister, and they have launched an AI tutor based on one of the GPT family APIs nationwide. This is fantastic. Additionally, I met with the founder who is building it because they are using various technologies to improve this model. The prospect of providing personalized tutors for students worldwide is absolutely achievable.

When you consider government transfers and education subsidies that exist all around the world, including in developing countries, being able to provide highly capable personalized tutors now, I think it is possible. So I am very excited about this. Another area is health, if you are interested. Even in the United States, 19% of our GDP is spent on healthcare. We know we need to improve health outcomes and reduce costs. And a lot of the costs are actually workflow costs, honestly, not even necessarily some magical drug or something like that. Those are great, and they are important. But the real cost is in care management.

That’s why I’m excited about partnering with Epic, because they can have a huge impact. Epic is everywhere, they can do anything. They are deeply building generative AI. I mean, they told me all these use cases as you finish your shift and you want to hand off the process. Oh, why not do a summary? Why not have a robot by the bedside that can ask all the questions and not miss anything?

For doctors’ inboxes, we now have this incredible way to really record the conversation the doctor has with the patient and reduce the burden on the doctor. So I think once applications and technologies are developed, once this software is built, it can go to every hospital in the world. So I think we can have a huge impact in these two areas.

Elad Gil:

We’ve talked a lot about AI. Obviously, Microsoft has done a lot in many important areas like B2B consumers. Besides AI, what other areas excite you?

Satya Nadella:

We just completed the acquisition of Activision Blizzard. We are very excited about gaming. You mentioned “Minecraft”. So gaming is another area that we consider as a category of products. In fact, if I look back at Microsoft, there are three things that, naturally for us, you could almost say our DNA. One, these platforms and tools, we are always going to be a platform development tools company. That’s core tradition. The second is productivity and communication software. I think that’s another thing we’re going to do. And third is gaming. In fact, I think Flight Simulator was built before Windows. So we will always be working in these three areas. Additionally, in the gaming space, what excites me is we love consoles, we love PC gaming, but with some of the changes with cloud gaming, we can take gaming anywhere. So I think that’s where AI as a platform in all of these areas can really make a impact.

Elad Gil:

I think your performance as CEO is one of the most successful in the history of capitalism, and I’m not saying this lightly. When you took over, the company had a market capitalization of $250 billion, which some people thought was a crazy high valuation. Now your market cap has reached $25 trillion. So in the past eight to nine years, you have added $22.5 trillion in market value to the company. What kind of legacy do you want to leave at Microsoft? How do you think about this question for the future? Or, looking back in a few years, what achievements would you like to have made?

Satya Nadella:

I’m very skeptical of views on legacy or other aspects. Anyone who says, “Oh, I took over and the person before me was terrible, and I changed everything and it’s all because of me,” I’m very skeptical of those people, because to some extent, the goal of work is to build the institution strong enough that the person who succeeds you is more successful than you. In fact, I always think, if the person who replaces me can be successful in any role, then maybe you’ve done something right.

Frankly, in the tech industry, there is no value in franchising. We know you have to constantly reinvent yourself. That means you have to ensure you leave enough energy in the system for it to continue renewing itself. And sometimes, measuring success by market cap is absolutely not the way to measure things, because you have to make long-term investments before traditional wisdom catches up. You can’t expect the market to always reward you. The market will reward you, but there will be a time, as Bezos said, where you have to be comfortable being misunderstood, and I think that’s right.

So I see all of this as long-term indicators or indicators of things going well, but not the only indicators. What I’m more focused on is ensuring Microsoft does something relevant in the future. If we’ve created enough institutional and cultural strength where we’re not doing things out of envy, but we’re doing things that are useful to the world. We’re doing things that help the company succeed while also creating, I think, an incredible balance if our world around us is doing well and we’re doing well.

Elad Gil:

Makes a lot of sense. So when you look at other important technology areas that people are talking about now, like fusion or other forms of energy, autonomous driving cars and how they change cities and transportation, there’s all sorts of areas. Are there any in particular that you’re most interested in or most curious about the societal impact they might bring?

Satya Nadella:

When we think about AI and what we’re building, we definitely need energy. In some sense, our fortunate position is that all the new projects start from the world’s largest green energy purchaser today, which means we can incentivize a lot of these new projects, sometimes with risks, but I think the most important thing we have to do is the energy transformation.

Energy transition is a complex process. I never realized this, in fact, I thought we were talking about compressing 250 years of chemical knowledge and the entire petrochemical value chain into 25 years, which is a real challenge.

Even from an AI perspective, I’m excited about how this AI technology can become very useful and somehow synthesize some digital therapies that contribute to new batteries, and then these new batteries help us harness abundant solar energy, allowing us to utilize it more effectively in data centers, right?

These are real problems that need to be solved quickly. So, compared to other industries, I believe this entire value chain has both adjacency and practical dependence on us.

We will continue to update Blocking; if you have any questions or suggestions, please contact us!