Source: Intelligent Investor

I truly believe that success itself brings challenges because one gradually forgets what led to that success.

Those who grow alongside the company often feel that success is brought by themselves, but the reality is that we all rise with the tide. When the tide recedes, we must rediscover the foundations of the company.

This kind of forgetting is often the reason large successful companies decline.

In March of this year, $Microsoft (MSFT.US)$ CEO Satya Nadella engaged in a profound dialogue about the future of the Technology industry during an interview on the South Park Commons podcast.

He reflected on his career, discussed how Microsoft is redefining competitiveness in the era of AI, and shared his thoughts on the future of computational architecture, quantum computing, and the direction of intelligent agent development.

On February 4, 2014, Nadella took over Microsoft. At that time, Microsoft was gradually falling behind due to missing a series of innovative opportunities amid the wave of information technology changes. $Apple (AAPL.US)$ 、 $Amazon (AMZN.US)$ 、 $Alphabet-C (GOOG.US)$ 、 $Meta Platforms (META.US)$ As other companies surpass, the market cap has fallen to 200 billion USD.

He redefined Microsoft's mission: "Empower every person and every organization on the planet to achieve more." He brought together fragmented teams and broke new ground, injecting Microsoft with new energy and a grand vision: AI, mixed reality, and quantum computing.

Today, Microsoft's market cap is nearly 2.9 trillion USD (as of March 21, 2024).

In this interview, Nadella shared his understanding of how businesses can achieve legacy and evolution in competition, emphasizing that choosing the right competitive track is more important than simply winning.

He mentioned that in the Technology industry, "there is no real moat," and innovative companies must continuously disrupt and evolve, or they will be left behind by the times. "We must keep 'shooting,' or we won't even have a chance to score."

Nadella believes that the future development of AI has three key directions, and Microsoft is deeply laying out around these trends. He particularly emphasizes that simple AI models can no longer support long-term competition; a complete computing architecture and successful products are the key.

He believes that the computing architecture in the AI era and how to effectively utilize computing resources on the client side are two key directions that are underestimated.

In the field of AI, Nadella believes that the real value of AI lies in changing the boundaries of "accessibility," allowing more people to gain technological empowerment. To dance with AI, we need to maintain curiosity and critical thinking.

He raised a question, in the AI era do we need to rethink the essence of Education? And provided an open-ended thought, "If future children can use AI to build anything they want, should we foster their spirit of exploration and confidence more, rather than overly emphasizing exams and traditional 'knowledge reserves'?"

In addition to AI, Nadella also delved into the future of quantum computing. Microsoft has invested in this field for over 20 years, but he believes that quantum computing will not replace classical computing, but rather develop in synergy with it.

The topics discussed in this conversation, especially how to redefine corporate competitiveness in the era of AI, Cloud Computing, and quantum computing, are worthy of deep consideration by the Technology and Investment Industry.

1. Reflecting on entering Microsoft: Realizing that trends are irreversible.

Host: We want to revisit the key moments in your career and see your thoughts and experiences during the 'Zero to One' phase. You joined Microsoft in 1992. What was it like to interview at Microsoft at that time? Why did you choose Microsoft rather than going to a Venture?

Nadella: The 'Zero to One' concept is quite interesting.

Thirty-three years ago, I was actually working at Sun Microsystems. My first job after graduating from graduate school was in Mountain View. In 1991, I attended a developer conference at the Moscone Center, where Microsoft first launched Windows NT and introduced Win32 for the first time.

At that time, I suddenly had an idea (of course looking back now, it might be a bit hindsight): the X86 architecture would win.

When platform transitions occur, they are often irreversible, and although it takes time, the trend is clear.

It made me realize that even though Sun Microsystems was incredibly powerful at the time with a full-stack system, designing its own chips, developing operating systems, and even venturing into office software, I still believed that X86 was the future.

This prompted me to join Microsoft, even though at the time I was neither a Windows user nor had I ever been particularly fond of DOS.

But I felt Microsoft was the company that could truly seize this trend, and more importantly, it could genuinely become a force driving the popularization of Technology.

Host: Was your interview process intense? You should know, Microsoft's interview is not easy.

Nadella: Haha, Microsoft's interviews back then were typical of the old school Microsoft style. I think the interview process now is a bit better than it was back then; at least you can practice LeetCode instead of directly solving problems on a blackboard or whiteboard.

That's how I interviewed back then, and somehow, I actually passed.

But I will never forget the last interview question, which was probably the worst answer I ever gave.

At that time, Microsoft had an interview segment called 'As Appropriate'. If you successfully passed the previous interviews, you would be scheduled to meet someone whose name was not on the interview list, going directly to their office. This usually meant that you were basically going to be hired.

And the last question from this person is still vivid in my memory. He asked me, 'Assuming you are crossing the street and a baby falls, what would you do?'

At that moment, I thought, what kind of algorithm question is this? Is it some kind of problem I haven't learned? Then I answered, 'I would call 911.'

As a result, he stood up, escorted me out of the office, and said to me, 'Do you know what you should do when a baby falls? You should pick him up and comfort him.'

At that time, I thought, that's it, I'm definitely out of the running.

But it turns out that life will always give you some unexpected learning experiences.

2. A good organizational culture can help with inheritance and evolution.

Host: In the 1990s, Microsoft virtually covered all key technology areas and established a diversified technology ecosystem. Although today's Microsoft culture is different from that time, how would you describe Microsoft's culture?

On April 4th this year, Microsoft will celebrate its 50th anniversary. Recently, we have also been discussing what the company's culture is and what its core philosophy is.

Bill Gates and Paul Allen had a metaphor, or rather an idea: they wanted to establish a "Software Factory." It should be noted that there was no such thing as a software industry before Microsoft was founded. Therefore, their goal was to create a company focused on Software Development rather than a specific category of Software.

Initially, they started with the BASIC interpreter for the Altair Computer, which was the most suitable starting point at the time.

In fact, Microsoft's Flight Simulator was released even earlier than Windows, making it one of our oldest products.

So, in a sense, Microsoft has been looking for various application scenarios of the Software Factory, which has even led the development of knowledge-based work.

Microsoft's culture is about how to gather the best talents to build the strongest Software Factory, thereby creating the software products that the world needs.

Looking back, I believe this way of thinking provides valuable insights for all Ventures.

I find that many companies become deeply obsessed with a particular product, technology, or business model. But the problem is, everything has a day of being exhausted. I am also contemplating whether this concept of a "Software Factory" has an impact on Microsoft's culture.

When considering company culture, it becomes clear that the key lies in whether the talents coming in can truly create new value for this "factory."

Whenever there are better ideas or more advanced technologies, is Microsoft able to bravely move forward instead of sticking to a product built last year? Perhaps this culture has helped us maintain our vitality.

I do not believe that culture is a static thing; we have been striving to cultivate a learning culture.

Nowadays, I like to use terms like "growth mindset."

But in reality, I'm wondering if Microsoft had this growth mindset and learning culture from the very beginning. This might be the impact of the company's original structure.

Host: Are you saying that "software factory" is a commonly used term within Microsoft?

Nadella: Not really. In the Microsoft I grew up in, we preferred to call ourselves a "software company."

This "factory" metaphor arose more while we were discussing the 50-year history, especially now, as we are on the forefront of another technological revolution.

Today, as AI and AI agents become trends, how do we redefine the 'Software Factory'? This is the question I am currently pondering.

I hope to find answers in Microsoft's history and make this concept more relevant today.

Merely living long is not the goal; maintaining inheritance and evolution is. So, what gives us the courage to continue inheriting and evolving in the future? This has been a subject of my research, learning, and contemplation, as well as something I wish to express clearly.

3. The essence of Technology is constant disruption, and we must keep 'shooting on goal'.

Host: From 2000 to 2010, Microsoft seemed to go through a period of 'lostness'. Before that, you had won almost everything, but then for a time, you were no longer the most dazzling winner in the market.

How did you feel at that time? How did the company's culture respond? How did you personally experience this period?

Nadella: In the 90s, the gap between Microsoft and its competitors was very significant. Especially in the late 90s, there were only Microsoft and its competitors in the market, unlike today where there are seven or eight top technology companies globally, with new competitors emerging almost every day.

For instance, OpenAI has proven that a brand new company can quickly grow into a leader in the industry. The vibrancy of the technology industry today is incredible.

Success brings challenges, especially after achieving success, it's easy to forget what made you successful.

I believe that in the 2000s, Microsoft was not without innovative ideas, nor without attempts at new things. However, many times, we missed opportunities because we were too immersed in the existing successful models or unwilling to explore ideas that were not convenient or complete.

Steve Ballmer once gave me very valuable advice; he always said, 'You must have a complete idea.'

This statement has benefited me greatly. Many times, we may have some new product ideas, but they are often not mature enough. The reason why truly successful entrepreneurs achieve success is that their ideas are complete. If your idea is not complete enough, it is very difficult for founders to succeed.

Inside large companies, what you need is a group of 're-founders' who, at different stages, can have complete concepts and drive the company forward continuously.

Many successful large companies make the mistake of thinking of themselves as the reason for their success, rather than acknowledging that they are just riding the wave. When the trend changes, companies need to reshape themselves, rather than becoming immersed in past glories. I think this is precisely why many large successful businesses decline.

Host: What impresses me most is that Microsoft is really willing to learn and try new things.

Nadella: Yes, Microsoft has been persistent in trying, sometimes maybe too stubborn. But in the technology industry, this persistence is necessary.

You must keep trying and keep attacking.

Because the network effects in the Technology Industry are extremely brutal; if you can harness it, it will give you a huge advantage; but if you are on the wrong side, it will become very painful.

So, we must keep shooting, as that may be the only way to maintain.

Host: This is exactly what we tell founders in the early stages of a Venture, but seeing it apply at such a large scale of companies is truly shocking.

Nadella: That's right. I think the good news for the founders present is that there really isn't a moat in the Technology Industry. In a sense, this makes the Technology Industry extremely frightening.

You may remain profitable financially, but if you want to maintain legacy and evolution, you must continuously build new products, not just keep up with trends, but proactively invent new things. The nature of this industry is constant disruption.

4. Always focus on creating win-win long-term cooperation.

Host: Many people call Microsoft the 'King of Comebacks', especially after you became CEO. To be honest, I thought Microsoft would become one of those slowly declining tech giants, but after you took office, everything started to change. You have talked about reshaping Microsoft's culture, emphasizing the concept of 're-founding' and maintaining legacy and evolution.

Let's go back to the 'pre-phase' when you became CEO—what was the first thing you did? How did you drive this transformation?

Nadella: In the era of Cloud Computing, the real credit does not belong to me, but to Steve Ballmer. At that time, I was still his direct subordinate, and he was the one making the key decisions.

In fact, I learned a lot from Ballmer about how to be an effective CEO.

He gave me the freedom to invest in Cloud Computing, even when Wall Street was not on board.

This is what a true CEO should do: challenge mainstream views and empower the team, allowing people within the company to drive change.

In a sense, Ballmer completely ignored these external evaluations; he had already achieved great success at Microsoft. But it was his leadership style that taught me how to manage the company.

Bill Gates' vision and strategic thinking have also had a profound impact on me.

Host: Your investment in OpenAI and many other AI companies is particularly meaningful for today’s entrepreneurs present here. How should they collaborate with Microsoft? How can they pitch themselves if they want to attract funding from your Fund M12?

Microsoft's investment record is legendary, yet it is rarely discussed publicly.

(Note: Microsoft M12 is the company's venture capital fund, focusing on investing in early to growth-stage technology startups, initially established in 2016. M12 is not a purely financial venture fund, but an extension of Microsoft's ecosystem, with investments mainly serving Microsoft's strategic goals, such as promoting the development of Azure Cloud Computing, AI, Saas, enterprise software, and other fields. A classic example is the investment in OpenAI.)

Nadella's collaboration with the startup ecosystem, whether through M12 or other means, is very important for Microsoft.

But strictly speaking, we are not just a pure investment institution.

For example, when we wanted to invest in Facebook back then, it was not because of investment return considerations, but because we hoped to establish a partnership in the advertising business.

The situation with OpenAI is slightly different. When we initially supported them, they were just a research lab. Now, OpenAI has developed into a successful product company.

Microsoft has been closely watching Sam and his team, and adjusting our collaboration as their vision evolves. For us, OpenAI is not only an investment target but also a crucial business partner.

There are two core elements in Microsoft's DNA: we are a platform company and also a partner company. Without partners, there would be no stability in the platform.

Whether it's a startup or a mature company, there has always been a search for partners that can establish long-term, stable, and mutually beneficial cooperative relationships with Microsoft.

In the early stages of my career, while in charge of Microsoft’s database business, there was collaboration with SAP to make SQL Server the underlying database for SAP. This cooperative model was extremely beneficial for both sides. Similar examples include the Wintel (Windows + Intel collaboration model).

Therefore, what I always focus on is the ability to create win-win long-term partnerships, which are not only advantageous to Microsoft but also crucial for the partners themselves.

5. Simple models are no longer sufficient to support competition.

Host: Your investment in OpenAI is increasing. But in the face of competitors like Google and Meta, who all have their own foundational models, do you feel that Microsoft is somewhat lacking in this respect? After all, there used to be a saying: "Real companies own their own semiconductor factories." Today, can it be said that "Real companies should have their own foundational models"?

Nadella: My view is that Microsoft is a full-stack systems company, and we hope to have complete system capabilities, with foundational models being one important component.

When we first collaborated with OpenAI, it was based on a very clear technical judgment, which is Scaling Laws.

At that time, Sam, Ilya, Greg, and others firmly believed that large-scale training models could lead to breakthrough results. In fact, my earliest exposure to this concept was through a paper on scaling laws published by Dario (Amodei), which made me realize that this would be the core transformative direction in the field of AI.

The bet I made at that time was that the focus of the core transformation was in the field of natural language. Without breakthroughs in natural language, we might not have pursued this.

Now I have a deeper understanding of reinforcement learning, but if OpenAI's direction had not been natural language from the beginning, we might not have reached today.

After all, as a company centered around knowledge workers, Microsoft has always been obsessed with natural language processing, which has also been one of Bill Gates' primary areas of interest over the years.

For a long time, our way of thinking was to understand the world in a structured way, that is, defining people, places, and things, building the most complete ontology and models, and then allowing intelligent systems to grow through this approach.

We have taken many detours on this path.

Eventually, I began to realize that this approach might not be suitable for the real world because the world is inherently too chaotic, and structured methods like SQL cannot fully adapt to it.

I have always thought that someone would find another way to build semantic understanding, and this time the bet was placed on OpenAI.

Initially, I wasn't sure whether it would work, but over time, we saw it starting to take effect.

Looking back now, the collaboration between Microsoft and OpenAI is a perfect business partnership. We have not only jointly built the underlying system but also developed tools and products around it.

In the long term, we have a stable cooperation framework with OpenAI, and we possess relevant intellectual property. Therefore, we have the capacity to build our products, just like Mustafa Suleyman and his team have done with Pi (Inflection AI), and we are able to do it internally at Microsoft.

At this stage, my view is that the model itself is turning into a 'commodity' within Cloud Computing.

In fact, OpenAI is no longer just a model company, but a product company that happens to have one of the best models in the world.

This is great for them; for us, as their partner, it is also a tremendous advantage.

Now I am more focused on the new structure forming within the Industry: mere models are insufficient to support competition; a complete system architecture and successful products are key.

6. There are three clear directions for future AI development.

Host: In the past two and a half years, ChatGPT has truly entered the public eye, and we have witnessed a technological tsunami, from chips and datacenters to foundational models, vertical AI companies, as well as consumer Hardware and enterprise Software, with the whole Industry experiencing significant changes.

You are positioned at the core of the Industry and can see the full scope of these changes. If you were to give advice to the Entrepreneurs present, what do you think are the two currently underestimated opportunity areas? Which directions should entrepreneurs bet on in the next five years?

Nadella: I believe there are three clear directions, all of which are currently developing in parallel, making the entire Industry extremely exciting.

Now, our focus is not just on knowledge workers or language models themselves, but on a broader context:

First, AI applications in the real world: how does AI truly affect physical operations and interactions in the real world, not just in the digital realm? This involves fields such as robots, automated production, and supply chain optimization.

Second, AI models in the scientific field: whether in chemistry, biology, or other fundamental science fields, AI models are becoming a new cornerstone of scientific research. Although these models have some technical overlap with language models, their development also has its independence.

Third, innovation in system architecture: what kind of evolution is needed in the underlying architecture of AI systems to support future computing demands?

One question that I am increasingly concerned about now is, what kind of system architecture can truly support future AI?

A legendary engineer at Microsoft, Paul Maritz, taught me something very important: you need to predict future computing workloads and then design the appropriate underlying system to support it.

If we carefully observe the current development trends, it can be seen that the computing architecture of the AI era is at a critical transformative node.

The next generation of computing architecture not only needs to adapt to existing AI applications, but also considers how AI will further penetrate into the real world and scientific research fields. These changes are redefining the landscape of the entire Technology industry.

It is believed that we have not truly redesigned the system architecture from the perspective of first principles. We are addicted to AI accelerators, stacking a large amount of computing resources into existing Datacenters, and then hoping they can "fend for themselves."

Although certain results have been achieved, it is now time to think: What should the next generation of Hyper-Converged Infrastructure look like?

How should we redefine the synergy between computing, storage, and AI accelerators? In distributed synchronous training tasks, after introducing reinforcement learning and automatic scoring mechanisms, how should the system operate?

Even in the Datacenter field, it is often said that "a Datacenter is like a computer," but have we reached the stage where we need to build its 2.0 version?

This is a direction that I believe has been overlooked.

Another overlooked direction is the client side.

Nowadays, everyone is talking about Hybrid AI, but we have not truly cracked the issue of model architecture. Why don't we have a distributed inference model architecture? I am not talking about distributed training, but distributed computation at runtime.

Why can't I get the 45 TOPS NPU (Neural Processing Unit) running on my Copilot PC to collaborate with the cloud?

Currently, we can run some auxiliary functions, but can we fundamentally rethink the model architecture? Perhaps this requires a mathematical breakthrough to achieve.

The open-source community has done a great job in improving the computational efficiency of existing infrastructure, such as the excellent work of DeepSeek.

But why has no one truly cracked how to utilize the abundant computing resources available on the edge? This is an important opportunity that has been overlooked.

Host: This is interesting; returning to your earlier point, it seems that very few people have actually formed a complete thinking framework. Everyone is building certain small components, but a holistic architecture has yet to be formed.

Nadella: Your summary is very accurate.

Another example is AI systems themselves; even under the architecture of Copilot, whether it's GitHub Copilot or Copilot for knowledge workers, they are not yet a complete computing system.

If we say that AI is the new Computer, then it has not yet had its "Store-Program Moment."

Today's AI systems lack the following core elements:

1. There is no truly robust storage system, especially an architecture that supports multimodal storage;

2. The toolchain is still too "craft-like," and the developer experience still needs improvement;

3. There is a lack of system-level integration; current AI products are more like scattered components rather than a coherent system.

Of course, excellent research is already underway in these areas, which is why it feels like we are still in the early stages of building AI systems.

It is reassuring that we are seeing not only more foundational models but also an increasing number of Ventures focusing on AI system architecture.

7. A true intelligent agent should be a collaborative partner capable of executing complex tasks.

Host: When we think about a complete system, I like to envision the world we hope to be in five years from now. How do you define 'true intelligent agent behavior'? Not just Copilot, but a broader concept of an intelligent agent?

Nadella: This is a question I have been thinking a lot about recently.

If someone had told me before the PC era that for most of my 33 years at Microsoft, I would be a typist, I would have found it unbelievable.

At that time, I would have thought, 'Computers will completely change my life, and they will allow me to do unprecedented things.'

But the reality is that one of the core applications of the PC indeed was typing. This made me realize that our expectations of AI agents today may also lead to similar misjudgments.

Therefore, when we talk about 'true agents', what I want is an AI that can actively reason, adapt to different environments, and can deeply interact with the real world.

It is not just a tool, but a collaborative partner capable of executing complex tasks.

How will this vision affect my daily life? It needs to possess the following characteristics:

1. Autonomous planning and execution. The AI can create its own plans instead of just waiting for human input;

2. Cross-modal capability. It needs to integrate various modalities such as vision, text, voice, and code, rather than just relying on natural language; 3. Personalized adaptation. It can continuously optimize its behavior based on the user's habits and preferences, rather than being a fixed "assistant" role.

A true AI agent should be a partner that helps us create, execute, and optimize tasks, rather than just an enhanced search engine or chat bot.

If this can be achieved, it will be the fourth revolution in the computer industry following PCs, the Internet, and mobile computing.

Every morning I wake up and turn on my computer to type, and then shut it down to sleep at night. So I wonder if my future job will become managing my AI agent's inbox.

Even in a world dominated by agents, I still need some degree of control. That's why I am so focused on Copilot as the user interface for AI.

Because no matter how I delegate tasks, the AI will ultimately come back to me to notify me, seek permission, receive instructions, and help me complete the tasks. Hopefully, the quality of these notifications will be more efficient than today's emails or messages.

In a sense, we need a new UI layer that lets AI truly become our leverage.

I imagined two metaphors:

Personal life: I hope AI will be my friend, coach, and advisor.

Work scenario: I hope AI will be my chief of staff, researcher, and consultant, helping me complete all work tasks.

I believe this will be my future relationship with AI, and I hope it can completely change the tedious parts of knowledge work.

Another question worth pondering is that we often confuse the concepts of 'knowledge work' and 'knowledge workers.' There will still be knowledge workers in the future, but the level of work they engage in will be different.

In the 1990s, Word, Excel, and PowerPoint were the core tools for knowledge workers; Copilot is their version 1.0, while the next generation of AI tools will be version 2.0.

So, what is version 3.0? That is precisely the direction we are striving to envision.

Host: If you envision the future AI user interface, what do you think it should look like? Today's chatbots or Copilot are still not intuitive enough; they lack true integration capabilities.

Nadella completely agrees. Therefore, I value multimodal interfaces very much.

My favorite way of interaction is the voice interface of Copilot. For example, I set up shortcuts on CarPlay on my iPhone to interact with AI using voice.

In the past, I listened to podcasts by playing them directly, but now, I prefer to have a conversation with the transcript of the podcast during my commute. AI can read the content aloud, and I can interrupt or ask questions at any time for it to summarize or explain key points.

This conversational capability is an unprecedented experience, completely transforming the way information is consumed.

It's like autocomplete; once you use it, there's no going back.

8. The future computing architecture will be 'Classical Computing + Quantum Computing.'

Host Bill Gates once said: 'People often overestimate the progress of technology in a year, but underestimate the changes in a decade.'

Last week, you just released some announcements about quantum computing. Does this technology currently have the conditions for implementation for Ventures, or is it still in the 'one-year phase,' far from large-scale applications?

Nadella discusses several aspects of quantum computing.

Firstly, Microsoft has been persistent in this field for over 20 years, and I might be the third CEO to fund quantum computing research.

Looking back at history, our thoughts have always been very clear: Bill Gates and Craig Mundy decided early on that we want to build a truly 'practical quantum computer'.

However, to achieve this, we first need to address the stability of quantum bits. This is one of the key challenges in the entire industry.

Currently, quantum computing is still in the research phase, but some early application cases are starting to emerge: in materials science and drug development, quantum computing has shown potential; the financial industry is exploring how to utilize quantum computing for more complex risk analysis; in the future, the combination of AI and quantum computing could lead to greater computational power.

For most Ventures, it may still be too early. Unless your Business directly involves quantum algorithms, quantum simulation, or materials computation, it is currently unlikely to establish a business model fully reliant on quantum computing.

But if you lay the groundwork in advance, on the day quantum computing truly matures, you will have a first-mover advantage.

To build stable quantum bits, we need to achieve breakthroughs in physics.

Microsoft has taken a unique approach in its research in this area, exploring a phenomenon known as Majorana Particles. These particles were theoretically predicted by a scientist in the 1930s, but there was no experimental evidence for decades.

After 20 years of research, we have finally proven the existence of Majorana Zero Modes. Moreover, we were able to fabricate these particles atom by atom and demonstrate that they can stably hide quantum information, thereby reducing the need for error correction.

This means we can truly scale it up into a practically operable quantum chip.

This is indeed an incredible breakthrough.

Now, the development of quantum computing has become easier to advance than it was in 2014, as we have addressed the most core challenges in physics.

Although we faced some setbacks in the process, it was truly inspiring to see the team persevere and ultimately achieve breakthroughs.

One thing I am increasingly clear about is that quantum computing will not replace classical computing; rather, it is their combination that will yield the greatest benefits.

Many people misunderstand quantum computing, believing it will ultimately replace classical computing entirely, but that is not the case.

For example: when researching Majorana particles, the approach used was to build atom by atom. If a Quantum Computer had been available at that time, the entire research process would have been much faster.

This is because the core advantage of Quantum Computers is simulating the complex phenomena of nature, which can help explore data spaces, but they are not suitable for data-intensive computational tasks.

This means that the future computational architecture will be a combination of 'Classical Computing + Quantum Computing':

Quantum Computing: good at exploring large data spaces, simulating physical, chemical, and biological processes; Classical Computing: good at processing vast amounts of data, performing AI reasoning, storage, and operations.

For instance, in scientific modeling, Quantum Computing can be used to generate training data, and then traditional AI models can be used for reasoning. This is exactly what Azure Quantum is currently doing.

Even now, with only 10 or 100 logical qubits, they can already play a role in the fields of chemistry and biology.

Because of this, we see more and more biopharmaceutical companies beginning to focus on Quantum Computing and willing to invest in this field.

Host: We must talk about the multiverse theory. This basically means that quantum events may point to the existence of many parallel universes. Do you believe in the multiverse theory.

Nadella: Haha, my viewpoint is closer to the Copenhagen interpretation rather than the multiverse theory.

But honestly, let the physicists debate these issues, I will focus on building a Software factory (laughs).

Host: But as a philosophical thinker, what is your perspective?

Nadella: My stance is that I exist, so there must be something that I can perceive, and these are measurable. I think this conclusion is reasonable enough.

9. Maintain curiosity and critical thinking, dance with AI.

Host: What advice would you give us parents to prepare for our children’s future?

Nadella: That’s a great question.

The core issue we are currently pondering is whether AI will enhance human autonomy or undermine it.

There are two possible directions:

AI enhances human capabilities, enabling humans to become more autonomous and focused on creative and strategic tasks;

AI undermines human autonomy, turning us into passive executors, subject to the decisions of algorithms and automated systems.

It is believed that AI will enhance our capabilities, provided that we properly guide it.

I have experienced two of the most impactful AI moments that profoundly made me understand how AI is changing the way humans work:

The first was seeing the early version of GitHub Copilot. When I first programmed with it, I thought to myself, “Wow, this thing can really write code!”

You have to understand that convincing a software engineer that AI code generation is valuable is one of the hardest things in the world (laughs).

The second was observing knowledge workers using Copilot. This made me realize that future knowledge workers will still exist, but their ways of working will change dramatically.

What we need to understand is that in the future, there will still be knowledge workers, but the knowledge work they engage in will be at different levels of abstraction.

I believe that future AI will become friends, coaches, and advisors in personal life, as well as chief staff, researchers, and consultants in the workplace. This is the future of AI that I hope to see.

We must teach children how to collaborate with AI, allowing AI to enhance their creativity, critical thinking, and independent decision-making skills, rather than having them replaced by AI.

This is our most important responsibility to the next generation.

In January 2022, I saw an application case in India, which was one of the most impactful AI demonstrations for me.

India has a complete set of open digital public infrastructure, which includes features like text-to-speech and speech-to-text, that can be used as public resources.

At that time, developers used these tools to build a WhatsApp-based AI assistant, enhancing its conversational capabilities by linking it to GPT-3.5.

A farmer in a rural area of India asked this AI assistant via WhatsApp: "I heard about a subsidy program on television. Can you help me check it out?"

The AI assistant replied: "You can visit this website and fill out the relevant form to apply."

The farmer continued: "I don't know what a form is, and I don't know how to access the website. Can you help me apply?"

The AI assistant directly helped him complete the application process.

This is still the GPT-3.5 version, but it has significantly lowered the technical barrier, allowing this farmer to access services that were previously unattainable.

This moment left me deeply shocked!

I realized that this farmer from India could now directly change his life using technology that was only developed a few months ago on the West Coast of the USA.

In the past, he was not incapable but lacked access to information and services. Now, AI has given him that capability, breaking the original limitations.

This change made me rethink what our children's future will look like.

If our children can code, research, and build whatever they want at any time, then the most important skill in future society may not be mastering a specific professional skill, but rather maintaining curiosity and critical thinking.

Do we need to rethink the essence of education.

If future children can use AI to build anything they want, should we focus more on cultivating their spirit of exploration and self-confidence, rather than overemphasizing exams and traditional notions of "knowledge reserve"?

Host Satya, you are clearly a highly competitive person, how do you unleash your competitive spirit outside of work? For example, do you pour this passion into cricket matches?

Nadella: Haha, I was indeed very happy with this morning's match! Like almost all South Asians, I am obsessed with cricket.

But recently, I have been thinking about the true meaning of competition.

Especially at this stage of my life, I increasingly realize: "I need to focus on the right competitive game, not just immerse myself in the competition itself."

I find that entrepreneurs inherently possess this ability; they need to define their own company and decide which competitive track they want to enter.

Therefore, they are always thinking: what kind of game do I really want to play?

The real challenge of competition is to ensure that you are in the 'right competitive landscape'. One unique aspect of the Technology industry is that the boundaries of competition are always changing.

Looking back at my experience, I came to the USA in 1988, the Berlin Wall fell in 1989, and I entered the Technology industry in 1990. Since then, I have experienced an incredible period of transformation.

This experience made me realize that many things happen simply by luck.

The key question is: how to seize the right opportunities in the correct competitive landscape? This is the key to determining success or failure.

The competitive landscape is the most important, this is becoming clearer to me.

You want to enter an industry with excellent competitors, rather than one where you might win, but it isn't truly great.

Therefore, my way of thinking has changed from 'playing games' to 'choosing the right game'. This makes me focus more on: what game are we really playing? Is this the competition we really want to engage in?

Competing in the 'Greatest Game' is a gift.

Being in the age of Computer and being able to participate in the greatest Technology competition of our generation is truly a blessing.

We are in an unprecedented era of innovation, and being part of it is truly a precious gift.

Editor/rice