Core ideas:

1. Facing the "trough period" when Nvidia plummeted by 80%, I returned to the "core" of things, insisted on what I believed in, and then moved on without changing anything.

2. When Nvidia was founded, its first major decision was to determine that 3D graphics technology would become the first "killer" application. The company's mission is to "build special computers to solve problems that ordinary computers cannot solve."

3. Almost all of NVIDIA’s work over the past thirty years has revolved around technology and the market. This is also the core of NVIDIA: create technology before creating demand. In the next ten years, NVIDIA's biggest challenges will come from technology and market, and other challenges will come from industry, geopolitics and social levels.

4. NVIDIA can make a unique contribution to advancing the future of computing, which is one of the most important tools for mankind.

5. The latest major breakthrough in AI is deep learning. Another important breakthrough is the newly invented reinforcement learning based on human feedback (RLHF) technology for language models. NVIDIA has found a solution to implement this technology at the system level.

6. In the future, the way information is processed will fundamentally change. Generative AI will start from an information "seed". The future of computing will be highly dependent on generation rather than retrieval.

7. The future supervision of AI will come from two aspects: social supervision and product and service supervision. The social problems brought about by the development of AI should be broken down and supervised.

8. With the development of technology, information processing and software development will undergo fundamental changes, leading to major adjustments in industry structure and organizational methods. Reshaping the organizational structure strives to abandon traditional hierarchical management and advocate a more equal and open organizational culture.

9. In a cultural environment that does not believe in "information = power", trust and empowerment of employees are very important. With only 30,000 employees, NVIDIA is the smallest large company in the world, but each employee is given tremendous power.

10. I hope that NVIDIA can be remembered by history as "changed everything" by persevering in doing what it is good at and passionate about.

Since last year, riding on the intensifying AI wave and holding "hard currency" GPUs, Nvidia has successfully become the "number one AI chip". The company's stock price has soared 300%, and its market value has reached US$2 trillion, making it one of the best in the United States. second.

On March 4, local time, Nvidia CEO Jen-Hsun Huang, whose net worth has reached the top 21 in the world, went to Stanford Business School and accepted an interview from the "View From The Top" series of salon events.

In this latest public interview, Huang Jen-Hsun still appeared in the classic "black leather jacket". As a graduate with a master's degree in electrical engineering 34 years ago, Huang Renxun, as a Stanford alumnus, shared how he founded NVIDIA, successfully obtained the first funding, how to establish the company's organizational structure, and how to build the first "killer" application. In addition, Huang Renxun also expressed his views on the artificial intelligence revolution and how to survive the "trough period".

Working hard to create NVIDIA

Q: Jensen, it's an honor to have you. Thank you for coming.

A: It's a pleasure to be here.

Q: Thank you. To commemorate your visit to Stanford again, let’s start from the moment you first left. You chose to join LSI Logic, a company that was very attractive at the time, and worked with top people in the technology world, but decided to leave and start your own business. What inspires you?

A: I was working as an engineer at LSI Logic, and Chris and Curtis were at Sun at the time. Working with some of the top names in computer science at the time, including Andy Bectoshim, we were involved in building workstations and graphics workstations. One day, Chris and Curtis expressed their wish to leave Sun and wanted me to start a business together. Even though I had a great job, they insisted on inviting me to co-found a company. So whenever they would visit, we would meet at Dennis, which, by the way, is my alma mater. I worked as a dishwasher, my first employment experience, and I did really well. We started brainstorming ideas at the time of the microprocessor revolution, in 1993 and 1992. The PC revolution was in its infancy, with Windows 95 not yet available and Pentium not yet announced. Before this PC revolution was about to come, the importance of the microprocessor was obvious.

We thought, why not start a company? The goal is to solve problems that cannot be solved by general-purpose computing. So that became the company's mission: to build special computers that could solve problems that ordinary computers couldn't. To this day, we remain focused on this. If you look at the markets we've expanded as a result, including computational drug design, weather simulation, materials design, etc., we're very proud of that. There’s also robotics, self-driving cars, automation software we call artificial intelligence, and more. We pushed technology to its limits, eventually making computing cost almost nothing, which led to a completely new approach to software development, where computers write their own software, which still amazes me to this day. This is our journey. Thank you all for coming.

Q: These apps are on all of our minds today. But at that time, the CEO of LSI Logic convinced his largest investor, Don Valentine, the founder of Sequoia (Sequoia Capital), to meet with you. I see a lot of founders out there who are eagerly waiting, but how do you convince Silicon Valley’s most sought-after investors to invest in a team of first-time entrepreneurs building a new product for a market that doesn’t exist yet?

A: I had no idea how to create a business plan, so I walked into the bookstore that still existed. In the business books section, I found a book by a familiar author - Gordon Bell. The title of this dauntingly thick book is "How to Write a Business Plan," which is both straightforward and yet extremely relevant. It seems that Gordon only wrote for a few people, and I accidentally became one of them. One should have known at the time of purchase that this was not a wise move; after all, Gordon's superior intellect meant he had much to teach.

When I picked up this 450-page book, I quickly realized that if I immersed myself in the book, I might end up with my company before I finished reading it. At the time, Laurie and I were only able to make ends meet for six months, and with the financial pressures of our children, Spencer and Madison, and a pet dog, I had to give up further reading. So I didn’t write a business plan. I just went to talk to Wilf Corrigan. He called me one day and said, hey, you left the company and you didn't even tell me what you were doing. I want you to come back and explain it to me. So I went back and explained this to Wilf. Wilf finally said, I don't know what you said. This is one of the worst elevator speeches I have ever heard. Then he picked up the phone and called Don Valentine. He called Don and he said, Don, I'm going to send a kid over there. I want you to give him money. He is one of the best employees LSI Logic has ever had.

What I learned from this experience is that you may have had successful interviews or experienced failed ones, but you cannot escape your past. Therefore, strive to create a good past.

I seriously say that I am the best dishwasher at Denny's. I plan my work well and stay organized. I'm the kind of person who keeps things organized. Then I worked really hard. After that, they promoted me to bus boy. I'm convinced I'm the best bus boy Denny's has ever had. I never leave my position idle, I never leave empty-handed, and I work extremely efficiently. So, eventually I became CEO. I'm still trying to be a good CEO.

In the face of headwinds, I stick to what I believe in

Q: With so much competition, it’s really hard to stand out among 89 companies chasing after building the same product. With only 69 months of operating capital remaining, you know your initial plans are no longer feasible. How do you decide your next steps to save your company in this kind of adversity?

A: We founded this company to advance the development of accelerated computing. But the key question is, for what purpose will this technology be used? What is its killer app? Our first major decision, and the reason why Sequoia Capital supported us, was to determine that the first killer application would be 3D graphics technology, and the application scenario would be video games. At the time, it was nearly impossible to create low-cost 3D graphics technology, with companies like Silicon Valley Graphics producing image generators that sold for millions of dollars. Moreover, the video game market at that time was almost zero. In this way, we face a huge challenge: how to combine a technology that is difficult to commoditize with a market that almost does not exist. It was this intersection that defined the founding of our company.

I still remember that at the end of my speech, Don gave me a profound inspiration. He said: "Startups should not invest in or cooperate with other startups." This means that Nvidia's success depends on On the success of another startup - Electronic Arts. As I left, he reminded me that Electronic Arts’ CTO was only 14 years old and still needed his mother to drive him to work. He wanted to use this story to remind me that our success depends on partners like this. He then half-jokingly warned me that if I lost his money, he would make me look good. This is my strongest memory of that meeting.

Still, we accomplished something. We spent several years creating the gaming market for PCs, and it was a long process that we're still continuing to this day. We realized that not only did we have to create technology, we had to invent a new way of doing computational graphics, and we had to turn a million-dollar technology into something that could be put into a computer for only three or four hundred dollars, and we had to create this whole new market. Therefore, we not only need to create technology, but also create markets. This innovative spirit defines NVIDIA.

Today, almost everything we do revolves around creating technologies and markets. That's why people say we have a technology stack, an ecosystem. This has been the core of NVIDIA for 30 years: In order to create the conditions for people to buy our products, we must invent new markets. That's why we're at the forefront of all these areas like autonomous driving, deep learning, and computational drug design and discovery, and we're creating the market as we're creating the technology.

Then, Microsoft introduced the Direct3D standard, which spawned hundreds of companies. After a few years, we found ourselves competing against just about everyone. The 3D graphics technology we invented is incompatible with Direct3D, which puts us in a very embarrassing situation. We were faced with a choice: reset the company or go out of business. But we don't know how to build it the Microsoft way. I remember in one weekend meeting we discussed that we now have 89 competitors. We know our way is wrong, but we don’t know what the right way is. Fortunately, we found books on the OpenGL pipeline at Fry's Electronics, which define how computers do graphics processing. I bought a few books to take back to the company and told my team that I had found our future. We implemented the OpenGL pipeline as described in the book and created something the world has never seen before. Many lessons learned during this process laid the foundation for our company’s subsequent success.

That moment gave us a lot of confidence. It proves that it is possible to successfully create the future even when you know nothing about something.

That's my approach to everything now. When someone introduces me to something I've never heard of, my first reaction is always: "How hard is this?" Often just a textbook or paper away. So I spent a lot of time reading research papers. And then, if that's true, you can certainly learn how others do something and then go back to the original principles and ask yourself, given today's conditions, my motivations, my tools, and how things have changed, how would I go about it again? Do this? How would I reinvent the whole thing? If I were building a car today, would I start with the 1950s and 1900s and make incremental improvements? How would I build a computer today? How would I write software today? In this way, I always go back to the original principles, even in today's company, to reposition myself because the world has changed. The way we used to write software was monolithic and it was designed for supercomputers, but now it's decentralized. Now how we think about software today, how we think about computers today, how we think about things, always bringing your company, bringing yourself back to the original principles, that creates a lot of opportunity.

If you send me something you want me to review, I'll do my best and show you how I'd handle it. In this process, I certainly learned a lot from you. Is it right? You gave me the seeds of a lot of information. I have learnt a lot. Therefore, this process has been beneficial to me. Sometimes it does take a lot of energy because, in order to add value to someone else, and they're very smart to begin with, and I'm surrounded by very smart people, you have to be at least on their level. You have to get into their headspace, which is really hard. This requires tremendous emotional and intellectual energy. So when I'm dealing with this kind of stuff, I get tired. I have a lot of great people around me.

In theory, the CEO should have the most direct reports because those who report to the CEO require the least management. It makes no sense to me that a CEO has very few direct reports, except for one fact that I know to be true. That is, the CEO's knowledge and information are considered extremely valuable and extremely confidential. You can only share it with two or three people. Their information is so priceless and confidential that they can only share it with a few more people.

I don’t believe in a culture and environment where having information gives you power. I want all of us to contribute to the company, and our place in the company should be tied to our ability to solve complex problems, guide others to achieve greatness, inspire, empower and support others. These are the reasons why the management team exists, to serve all the other employees in the company, to create the conditions for all these amazing people to volunteer to come and work for you. Compared with all the annual high-tech companies in the world, they choose to work for you voluntarily. So you should create the conditions so they can do their life's work, and that's my mission. You may have heard that, I've said it very clearly. I see my job as very simply creating the conditions that allow you to do your life's work. So, how do I do this? What does this condition look like? What should be the consequences of this condition? Great empowerment.

You can only be empowered if you understand the situation, right? Right? You have to understand the context of your situation so that you can come up with great ideas. So I have to create an environment where you understand the context, which means you have to be informed. And the best way to be informed is to have as few levels of information distortion between us as possible.

That's why I often reason about things in settings like this or in front of an audience. Let me start by saying that these are the initial facts that we have. This is the data we have. Here's how I would reason. These are some assumptions, these are some unknowns, these are some knowns. So, you reason, now you've created a highly empowered organization. NVIDIA has 30,000 employees and is the smallest large company in the world. We are a small company, but every employee is empowered to make smart decisions on my behalf every day.

This is because they understand my position. I'm very transparent with people. I think I can trust you with the information.

Now, a lot of times the information is hard to hear and the situations are complicated, but I believe you can handle it. You, you know, a lot of you, listen to me, you're all adults. You guys can handle this. Sometimes they're not really adults. They just graduated. I'm just kidding.

I know that when I first graduated, I was barely an adult. But me, I'm lucky enough to be trusted with important information. So I wanted to do something like that. I want to create conditions for people.

The core of NVIDIA: pioneering a new way of computing

Q: Yes, the way you are applying this technology is proving to be revolutionary. You got all the momentum you needed to do an IPO because over the next four years, your revenue grew ninefold. But in the midst of all this success, you decided to shift Nvidia's innovation focus a little, based on a phone conversation you had with a chemistry professor. Can you tell us about the content of that phone call and how you connected what you heard to your next actions?

A: As I recall, the core of the company was to pioneer a new way of computing. Computer graphics was the first application, but we always knew there would be others. So image processing comes, particle physics comes, fluid dynamics comes, etc., all these interesting things we want to do. We make processors more programmable so that we can express more algorithms. Then one day we invented programmable shaders, which made all forms of imaging and computer graphics programmable. This is a great breakthrough. So we invented it. Based on this, we try to find ways to express more complex algorithms that can be computed on our processors, which are very different from CPUs. So we created this thing called CG. I think it was around 2003, and the C stood for GPU. It predates CUDA by about three years.

The same guy who saved the company, Mark Kilguard, wrote that textbook. CG is very cool. We wrote a textbook and started teaching people how to use it. We developed tools, rules, etc. Then some researchers discovered it. Many graduate students at Stanford use it. A lot of people who later became NVIDIA engineers were playing with it. I, some doctors at Massachusetts General Hospital picked it up and used it for CT reconstructions. So I flew over and looked at them and said, what are you doing? They told me. A computational quantum chemist then uses it to express his algorithm. So I realize there might be some evidence that people might want to use this. This gives us more confidence that we have to do this, that this form of computing can solve problems that ordinary computers really can't solve, and strengthens our belief to keep going.

Whenever you hear something new, you seem to cherish that surprise. This seems to be a theme in your leadership role at Nvidia. It seems like you made your bet long before the technological tipping point, and when the apple fell from the tree, you were standing there in your black leather jacket ready to catch it. How did you do it?

Always looks like a diving catch. You do things based on your core beliefs. We were convinced that we could create a computer that could solve problems that ordinary processors couldn't. CPUs are limited in what they can do, and general computing is limited in what they can do. And then there are some interesting problems that we can solve. The question is always, are these interesting problems just interesting, or can they also be interesting markets? Because if they're not interesting markets, it's not sustainable. NVIDIA went through about a decade where we were investing in this future and the market didn't exist. There was only one market at that time, and that was computer graphics, for 10 to 15 years. The market driving Nvidia today simply doesn't exist.

Well, when you have all the people around you, the management team at our company and NVIDIA and all the amazing engineers who are working with me to create this future, your shareholders, your board of directors, all of your partners, you take everyone with you Moving forward while the evidence from the market is not there is the real challenge. The fact that technology can solve a problem, and therefore the research papers that are possible, is interesting. But you're always looking for that market. But before the market exists, you still need early indicators of future success. Our company has a saying called key performance indicators (KPI). Unfortunately, KPIs are difficult to understand. I find KPIs difficult to understand and what is a good KPI. A lot of people, when we look for KPIs, you look at gross profit margin. That's not a KPI, that's a result. You're looking for something that's an early indication of a positive outcome in the future, okay? The reason is because you want early, early signs that you are going in the right direction. Therefore, we have this term called EIOFS, early indicators of success. It's helpful to people because I've been using it to give companies hope that, hey, look, we solved this problem, we solved that problem, we solved this problem. The market doesn't exist, but these are important questions, and that's what the company is about. To solve these problems, we want to develop sustainably. Therefore, the market must ultimately exist.

But you want to separate the results from the evidence that you're doing the right thing, okay? That's how you solve the problem of investing in something that's very far out and having the conviction to keep going, defining indicators as early as possible that you're doing the right thing. So start with a core belief. Unless something changes your mind, you continue to believe it. And look for early indicators of future success.

Q: What early success metrics do NVIDIA’s product teams use?

Answer: Various. Me, before I saw the paper, I met some people who needed my help, and they were doing this thing called deep learning. At that time, I didn’t even know what deep learning was. They need us to create a domain-specific language so that all their algorithms can be easily expressed on our processors. We created this thing called cuDNN, which is essentially SQL for storage computing. This is neural network computing. We created a language for this, the OpenGL of deep learning, if you will. So we, they need us to do this so they can express their math, and they don't understand CUDA, but they understand their deep learning. So we created this thing in the middle for them.

We do this because even though these researchers don't have money, these researchers don't have money. That's a great skill in our company, you're willing to do something even though the financial return is completely non-existent or maybe very remote, and even if you believe in it, we ask ourselves, is this work worth doing? Does this advance an important field of science? Note, and this is what I've been talking about from the beginning, we don't find inspiration in the size of the market, we find inspiration in the importance of the work. Because job importance is an early indicator of future markets. No one needs to do a business case, no one needs to show me a profit and loss statement, no one needs to show me financial projections.

The only question is, is this important work? If we don't do it, will it happen without us? Now, what if we don't do something? And without us, things still happen. This actually brings me great joy. The reason is, you can imagine the world becoming a better place. You don't even need to raise your hand. This is the definition of ultimate laziness. In many ways, you wish you had this habit. The reason is, you want the company to be lazy about things that others can always do. If someone else can do it, let them do it. We should choose things that if we don’t do, the world will collapse. You have to convince yourself that if I don't do this, it won't be done. This is important. If the work is hard and far-reaching and important, it will give you a sense of purpose. Does it make sense? Therefore, our company has been selecting these projects. Deep learning is just one of them. And the first sign of success was the fuzzy cat proposed by Andrew Anne. Then Alex Kerchewski detected cats, not every time successfully, but successfully enough to feel like, this might lead us somewhere. We reason about the architecture of deep learning. We are computer scientists and we know how to make things work. So we convince ourselves that this might change everything. Anyway, but that's just an example.

Q: So the choices you made paid off hugely, both literally and figuratively. But you have to guide the company through some very challenging times, like during the financial crisis when the company's market value dropped 80% because Wall Street didn't believe in your bet on machine learning. In times like these, how do you guide your company and keep employees motivated about the tasks at hand?

A: It was, my reaction in that period the same as my reaction to this week. You, earlier today, you asked me about this week. My pulse is exactly the same. This week is no different from last week or the weeks before that.

When your stock price drops 80%, you just want to wear a T-shirt that says, Not My Fault. But more importantly, you just don't want to leave your bed. You don't want to leave your house. All of this is true.

But then you go back to doing your job. Waking up the same way at the same time, prioritizing my day the same way, getting back to what I believe in. You have to core check, always come back to the core. What are the most important things and check them one by one. Sometimes it helps: Family loves me, OK, check; double check, right? So you just check it off and get back to your core and get back to work.

Then every conversation goes back to the core and allows the company to focus on the core. Do you believe it? Something has changed, stock prices have changed, but what else has changed? Have the laws of physics changed, have gravity changed? Of all the things we assumed, what we believed, that led to our decisions, what have changed? Because if those things change, you have to change everything. But if none of these things change, you change nothing. keep going. Yes, that's how you do it.

Q: When talking to your employees, they said you tried.

A: Avoid the public, including employees.

I'm just scared. Unfortunately leaders have to be seen. This is the hard part. I am a student of electrical engineering and I was very young when I went to school. I was only 16 years old when I went to college. So I was young in everything I did. Therefore, I am a bit introverted and a bit shy.

I don't like public speaking. I'm happy to be here. I'm not saying that, but it's not something I was born to do. And, when the situation is challenging, it's not easy to stand in front of the people you care about most. The reason is, can you imagine a company meeting and it's our stock price down 80%. And the most important thing for me as CEO is to come face to face with you and explain it. Partly you're not sure why, partly you're not sure how bad it's going to be, you just don't know these things. But you still have to explain it, to all these people, and you know what they're thinking. Some may think we are doomed. Some people may think you're a fool. Some may be thinking of other things. There's a lot of things that people are thinking, and you know they're thinking those things, but you still have to get in front of them and deal with it and do the hard work.

Q: They may be thinking about those things, but at a time like this, no one on your leadership team leaves.

A: That's what I keep reminding them. I'm just kidding. I'm surrounded by talented people. I'm surrounded by talented people. Yes, other geniuses. Nvidia is known to have the best management team in the world. This is the deepest technical management team in the world. I was surrounded by a lot of them and they were all geniuses. The commercial team, the marketing team, the sales team are incredible.

Empowerment is important in business organizations

Q: Incredible. Your employees say your leadership style is very participative. You have 50 direct reports. You encourage people from all parts of the organization to send you the top five things that come to mind, and you constantly remind people that no task is below you. Can you tell us why you intentionally designed such a flat organization? How should we think when designing the organizations of the future?

A: No task is below me. Because remember, I used to be a dishwasher, and I mean it. I used to clean toilets. I'm serious, I've cleaned more toilets than all of you combined. Some toilets are not visible at all. I don't know what to tell you. You know, that's life. Therefore, you cannot present me with a task that is below my capabilities. I don't do it just because, if it's below me.

If you send me something and you want my input on it, I can offer you a service by sharing with you how I reasoned as I review it. I made a contribution to you. I enable you to see how I reason about something. Understanding how someone reasoned about something, through reasoning, gives you power. He said, oh my gosh, this is how you reason about things like this. It's not as complicated as it seems. This is how you reason about very vague things. This is how you reason about things that cannot be calculated. This is how you reason about things that seem so scary. This is what you look like, you know? So I've been showing people how to reason about things. Strategic things, you know, how to predict something, how to break down a problem. You just empower people everywhere. So that's the way I look at it.

The future of computing will rely heavily on generation rather than retrieval

Q: Now let’s talk about the topic that everyone is most concerned about – AI. Last week, you mentioned that generative AI and accelerated computing have reached a tipping point. As this technology moves into the mainstream, what applications are you most looking forward to?

A: We need to go back to basics and think about what generative AI is. Through extensive learning and data analysis, we are now able to understand the meaning behind this data. Not only do we understand the meaning of various forms, we are also able to convert between them. This transformation is essentially the generation of information. This means that in the future, the way information is processed will fundamentally change, and the way software development and applications are processed will also change accordingly.

Previously, we relied on retrieval-based models, where information was pre-recorded and then retrieved based on algorithms. But in the future, generative AI will start from a seed of information and generate more content through prompts, making the future of computing highly dependent on generation rather than retrieval.

For example, in our current conversations, most information is generated on the fly rather than simply retroactively. This is intelligence. In the future, our computers will operate largely in this generative way, rather than relying on retrieval.

For entrepreneurs, this shift means rethinking which industries will be disrupted, how we think about networking, storage, and how our use of internet traffic will change.

From an organizational structure perspective, we should structure the organization based on fundamental principles. If we make different things, why do our organizational structures have to be exactly the same? Whether we are building computers or providing medical services, our organizational structure should reflect the characteristics of our environment.

Q: I would like to leave some time for the audience to ask questions. Your theme this year is "Redefining Tomorrow." The question we posed to all our guests was this: As the co-founder and CEO of NVIDIA, if you could close your eyes and magically change one thing about tomorrow, what would it be?

Q: Should we have considered this before? I might give an unsatisfactory answer. I think it's not just one thing. There are many things beyond our control. Your mission is to make a unique contribution, to live a meaningful life, to do something that no one else in the world can do so that when you're gone people will say that the world is a better place because of you. This is how I live.

Q: I always try to imagine the future first and then look back. Your question is exactly the opposite of how I think. I never look forward from my current position but instead imagine myself in the future and then look back. It's easier to do this. Looking back on the past is like reading your own history. We did this, we did that, we overcame this problem. Is this understandable? It's like the way you solve problems, first identify the outcome you want and then work backwards to achieve it. I have a dream that NVIDIA can make a unique contribution to advancing the future of computing, one of the most important tools for humanity.

It’s not that we have any respect for ourselves, but it’s something we’re really good at and it’s very difficult to do. We believe we can make an absolutely unique contribution. It has taken us 31 years to get here, but our journey has just begun. This is extremely difficult. When I look back, I believe we will be remembered as a company that changed everything, not by what we said, but by our persistence in doing something we were very good at and loved.

NVIDIA faces two challenges in the next 10 years

Question: I am a student graduating in 2023. My question is, how do you see your company developing over the next ten years, what challenges do you think the company will face, and how are you preparing for it?

A: First, let me talk about the thoughts that go through my mind. When you mentioned challenges, I had such a long list of challenges that came to mind and I was wondering which one to choose. To be honest, your question makes me think mostly of technical challenges, since that's my topic this morning. If you asked me yesterday, it would probably be a market creation challenge. There are some markets that I really want to explore. But we cannot do this alone.

NVIDIA is a technology platform company, and we serve many other companies, helping them realize the hopes and dreams we entrust through them. I hope that the biology community can reach a level similar to that of the chip design community 40 years ago, when computer-aided design (EDA) paved the way for everything we can do today. I believe we will enable computer-aided drug design for them. We can now represent genes, proteins, and even cells, and can almost understand the meaning of a cell. What does a cell mean? If we could understand a cell the way we understand a paragraph, imagine what we could do. So I'm anxious and excited about it. There are some technologies that I'm particularly looking forward to, like humanoid robots, that we've already turned the corner on. If we can understand language, why can't we understand operations? Once you solve one problem, you ask yourself, why can't we solve another problem? I'm very excited about these developments and it's been an enjoyable challenge.

Of course, the challenges we face are not just technical but also industrial, geopolitical and social. For example, global social and geopolitical issues, why can't we get along well with each other? Why do we magnify these problems? Why do we judge others so much in this world? You all know this already, I don't need to repeat it.

Expectations and worries about AI development coexist

Q: I am Jose, a 2023 business school graduate. My question is, are you concerned about the speed at which we are developing AI? Do you think some form of regulation is needed? Thanks.

Answer: Yes, the answer to the question is yes and no. We do need you. One of the recent breakthroughs in AI is deep learning, which has driven tremendous progress. Another important breakthrough is our newly invented reinforcement learning based on human feedback for language models. I practice this kind of reinforcement learning every day. That's my job, and for the parents out there, you're constantly providing that learning feedback.

Now, we have figured out how to achieve this at a system level for artificial intelligence. There are many other techniques that are necessary to set up safeguards, fine-tune, and ensure that the system conforms to actual physical laws. For example, how do we generate code names that comply with the laws of physics? Currently, some codenames seem to be floating in space and do not obey the laws of physics. This requires technology to achieve.

In terms of supervision, there are two types: social supervision and product and service supervision. I don’t know what to do about social supervision, but for product and service supervision, we know very well what should be done. Agencies such as the FDA and NHTSA have developed regulations for products and services for specific uses. Please don’t introduce a super-regulation across all sectors. The agency responsible for accounting regulation should not be regulating doctors.

But I overlooked an important question: How to deal with the social impact of AI? While I don't have a great question, enough people are talking about it. It’s important to break these issues down so that we don’t end up doing harm by focusing so much on one thing that we lose sight of the many regular things we could be doing. We should make sure we do the right things there.

The world’s most indispensable people in black leather jackets

Q: We also have some quick questions for you as usual.

A: Well, I was trying to avoid this. OK, let's get started.

Q: Your first job was at Denny's. They now have a dedicated booth honoring you. What is your fondest memory? Also, what’s the second job?

A: By the way, my second job was at AMD, is there a booth for me there too? I was just kidding. I really enjoyed my job there. AMD is a great company.

Q: If there was a shortage of black leather jackets around the world, what would we see you wearing?

A: Oh, don't worry, I have plenty in stock. I'll be the only one who doesn't have to worry.

Q: If you were to write a textbook, what would it be called?

Answer: Write a book? You're asking an impossible question.

Q: Finally, if you could share one piece of advice for Stanford students, what would it be?

A: It’s not a word, but I would say it is: having a core belief. Test it every day, pursue it with all your might, and stick with it for the long haul. Surround yourself with the people you love and enjoy the journey together. This is the NVIDIA story.

Q: Jensen, the past hour has been a blast. thank you for your sharing. Thank you so much.