NVIDIA CEO Jensen Huang's speech at the 130th Caltech commencement on June 14.

Today, NVIDIA has become the world's highest-valued company with a market capitalization of $3.34 trillion. (Although it has nothing to do with me because I am too poor to own any NVIDIA stocks; or perhaps because I don't have NVIDIA stocks, I am very poor 🐶), but coincidentally, today I had time to listen to the speech given by Huang on June 14 at the Caltech graduation ceremony recommended by Paula, which is also fitting for the occasion.

The general content of the speech:

Firstly, some polite remarks praising Caltech:

Richard Feynman, Linus Pauling, Carver Mead—these people who have profoundly influenced the chip industry and Huang himself graduated from Caltech.

Mr. Huang mentioned that the graduates' journey at Caltech symbolizes the sacrifices they have made for their dreams, as well as the character, determination, and willingness they have shown. Life also requires the ability to endure pain and torment. Most of NVIDIA's chief scientists graduated from Caltech, and he welcomes everyone to join NVIDIA. Today is a highlight moment for both NVIDIA and Caltech graduates, and the graduates will welcome more peaks in the future. Mr. Huang stated that he would continue to work hard, hoping for more glorious moments for the company in the future.

Mr. Huang is currently the longest-serving CEO of a tech company, having served for 31 years, striving to keep the company from going bankrupt, himself from getting bored, and not getting fired. As the CEO and founder, developing NVIDIA from zero to what it is today is itself a privilege.

reviewed the content of a speech given one year ago at National Taiwan University:

• CUDA was invented 20 years ago, which is today's revolution in computing.
• mentioned the Sega game console project.
• talked about intellectual honesty, as Richard Feynman said, intellectual honesty and humility saved the company.
• Lord Huang encouraged graduates to get involved in AI, which is the most important technology of our time. The results will be surprising, some are magical, some are disappointing, but you must enjoy it, you must participate in it, because it progresses very fast. It is the only technology that Lord Huang knows of that grows on multiple exponentials simultaneously. So, technological change is very, very fast. Therefore, Lord Huang advised the students of National Taiwan University to run, not just walk slowly, and actively engage in the AI revolution.

A share at National Taiwan University a year ago, and a share at Caltech a year later:

In one year, a lot has changed in the AI industry. The computer industry is undergoing a transformation from the ground up, really starting from the foundation. Every layer is changing, and it's happening very quickly; every industry will be transformed. The reason is obvious, because today's computers are the most important tool for knowledge. It is the foundation of every industry and every scientific field, and it will impact every industry.

The history of NVIDIA's GPU and AI was narrated:

• Modern computers can be traced back to the IBM System 360, whose basic ideas, architecture, and strategy still play a role in today's computer industry.
• In the 80s, Master Huang was one of the first-generation VLSI engineers who learned chip design from the groundbreaking textbook by Mead and Conway. Based on Carver Mead's pioneering work in chip design methodology at Caltech, this textbook revolutionized integrated circuit design. It enabled Master Huang's generation to design super-large chips, eventually leading to the development of CPUs.
• The CPU led to exponential growth in computing. Its performance and incredible technological advances, known as Moore's Law, drove the information technology revolution. Mass production of something the world had never seen before—mass production of invisible, easily replicable software—led to a $3 trillion industry. What was once considered a fantasy—that you could make money selling software—is today one of the most important goods, technologies, and products our industry produces.
• However, the limitations of NARD scaling, transistor scaling, and instruction-level parallelism have slowed the improvement of CPU performance. While computational demands continue to grow exponentially, the enhancement of CPU performance has slowed down. If the rapidly widening gap between computational demand and capability is not addressed, the rising energy consumption and cost of computation will eventually stifle industries across the board.
• NVIDIA's accelerated computing offers a way forward. By offloading time-consuming algorithms to GPUs specialized in parallel processing, we achieve accelerations of 10x, 100x, or even 1000x, saving money, costs, and energy. The applications we now accelerate range from computer graphics and ray tracing to gene sequencing, scientific computing, astronomy, quantum circuit simulation, SQL data processing, and even Pandas in data science.
• The hundredfold savings in time, cost, or energy through accelerated computing inevitably lead to new developments elsewhere. However, we didn't know what that would be until deep learning came along.
• Jeff Hinton, Alex Krizhevsky, and Ilya Sutskever trained AlexNet using NVIDIA CUDA GPUs and shocked the computer vision world at the 2012 ImageNet Challenge—this was the beginning of deep learning.
• In 2016, NVIDIA launched its first AI supercomputer, the DGX-1, and delivered the first unit to a startup in San Francisco that no one knew about, but it was a group of friends researching artificial intelligence called OpenAI.
• In 2022, ten years later, after a million-fold increase in compute capability, OpenAI released ChatGPT, bringing AI into the mainstream.
• During this decade, NVIDIA transformed from a graphics company initially known for making GPUs into an AI company that builds large-scale data center supercomputers.
• NVIDIA completely reinvented the company and also completely reinvented computing. The way we compute has fundamentally changed today.

Today with AI and predictions for the future:

• Today's compute stack uses GPUs to process large language models trained on supercomputers, rather than CPUs executing instructions written by programmers. We're creating software that no one person could write, doing things that were unimaginable a decade ago.
• Computers are now intent-driven rather than instruction-driven. Tell the computer what you want, and it will figure out how to do it. Like humans, AI applications will understand tasks, reason, plan, and coordinate a set of large language models to execute them.
• Future applications will operate in a way very similar to how we accomplish tasks—forming teams of experts, using tools, reasoning and planning, and executing tasks. Software and its capabilities have been completely transformed.
• Our industry has created an entirely new industry in this transformation. The inputs and outputs of AI are tokens, which are floating-point numbers embedded with intelligence.
• The company is now building a new type of data center that has never existed before, specifically for producing intelligent tokens. Essentially, it's an AI factory.

Compared the AI revolution to the electricity-driven industrial revolution:

Just as Nikola Tesla's invention of the AC generator drove the industrial revolution of the past, we now have AI token generators, and they will become the factories of the new industrial revolution. In the past, large-scale industries produced energy and electricity; now we have large-scale industries producing invisible software. In the near future, we will have industries producing intelligent tokens and AI generators. A new computing model has emerged, and a new industry has been formed, all because NVIDIA started from first principles, formed a belief in the future, and took action.

The next stage of the AI wave

The next wave of AI is in the field of robotics, where AI not only has language models but also physical world models. NVIDIA is working with hundreds of companies to create robots, robotic vehicles, pick-and-place robotic arms, humanoid robots, and even entire massive robotic warehouses.

However, unlike NVIDIA's AI factory strategy and experience where AI is formed through reasoning and deliberate actions, NVIDIA's journey into robotics has been marked by a series of setbacks:

• In 2000, NVIDIA invented the GPU and programmable shading, and launched an integrated graphics chip targeted at AMD CPUs. AMD wanted to control all the technology in PCs, while NVIDIA wished to remain independent. As a result, AMD acquired ATI and abandoned NVIDIA.
• NVIDIA turned to Intel and negotiated a license to connect to Intel CPUs. Apple was excited about what NVIDIA was developing and asked NVIDIA to co-develop the first-generation MacBook Air. However, upon seeing this, Intel became unhappy and terminated their agreement with NVIDIA.
• NVIDIA obtained a license from ARM and developed a low-power SoC, a mobile SoC that was a complete computer. The NVIDIA chip caught Google's interest, leading to a collaboration on a new device that eventually became the Android mobile device. However, Qualcomm was displeased and refused to allow NVIDIA to connect to their modem. Since there were no other LTE modem companies at the time, NVIDIA had to exit the mobile device market.
• With no market left to turn to, NVIDIA decided to develop something for which they were certain there would be no customers, because in one place you can be sure, there are no competitors where there are no customers, and no one will pay attention to you. Thus, NVIDIA chose a market without customers, a market worth $0, which was robotics.
• NVIDIA developed the world's first robot computer, processing an algorithm that no one understood at the time, called deep learning.

The learning in this process

This was more than 10 years ago, and Mr. Huang is extremely satisfied with what NVIDIA has developed and the opportunity to create the next wave of AI. More importantly, NVIDIA has cultivated a culture of agility and resilience. Time and time again, after setbacks, NVIDIA shakes off the dust and looks for the next opportunity.

Each time, it gains skills and strengthens the character of the company. It's hard to distract NVIDIA, and hard to discourage it. Now, any setback seems like an opportunity to Mr. Huang. Ironically, the robot computers NVIDIA develops today don't even need graphics processing, which was the reason NVIDIA's journey began.

The world is uncertain and the world can be unfair and deal you with tough cards.

-- Richard Feynman

However, there is always another opportunity, or an opportunity can be created.

A Short Story

On a hot summer day, Master Huang's family visited the Silver Pavilion in Kyoto. Master Huang noticed a gardener working alone, squatting there and carefully picking moss with bamboo tweezers, placing it into a bamboo basket. The basket looked almost empty. So Master Huang went over and asked him: "What are you doing?" He replied in English: "I'm picking the withered moss; I'm taking care of my garden." Master Huang said: "But your garden is so big." He answered: "I've been taking care of my garden for 25 years. I have all the time in the world."

This was one of the most profound lessons in Master Huang's life. It taught Master Huang: when you focus on your own career, you will find that you have all the time in the world.

A Day in Master Huang's Life

Master Huang starts every morning in the same way: by tackling the most important work first. Master Huang has a very clear list of priorities, starting each day with the highest priority tasks. Even before arriving at the office, Master Huang's day can already be considered successful because the most important work has already been completed, allowing the rest of the day to focus on helping others. When someone apologizes for interrupting, Master Huang always says: "I have all the time in the world," and indeed it is true.

The final piece of advice

I hope that each of you believes in something unusual, unexplored, but with evidence and rationality. Then commit yourself wholeheartedly to making it happen. You might find your GPU, your CUDA, your generative AI, your NVIDIA. I hope you see setbacks as new opportunities. Your pain and struggles will enhance your character, resilience, and agility; they are your most powerful superpowers.

Among all the abilities that Master Huang values the most, intelligence is not the most important. What Master Huang values the most is the ability to endure pain and suffering, the ability to dedicate oneself to something for a long time, and the ability to handle setbacks and see opportunities around the corner. Master Huang considers these his superpowers and hopes they can be your superpowers too.

I hope you find a craft. There's no need to decide on the first day, or even quickly, but I hope you find a craft that you're willing to perfect and hone for a lifetime, and let it become your lifelong career.

Finally, prioritize your life. There are many things and tasks, but deal with the important ones first so that you have enough time to do what matters.