Written by: Techub News Compilation

Introduction

Recently, NVIDIA founder and CEO Jensen Huang participated in an in-depth interview with Cleo Abram, host of the tech video channel "Huge If True." In a lengthy hour-long conversation, Jensen Huang unusually did not discuss the company's finances or management style but focused on the technology itself, aiming to "co-produce an explanatory video." He systematically elaborated on how NVIDIA started from gaming graphics cards and completely reshaped the modern computing paradigm through the idea of "accelerated computing," leading the next revolution in fields such as artificial intelligence, robotics, and digital biology. This conversation is not only a window into NVIDIA's technological philosophy and core beliefs but also a key guide to understanding technological trends in the next decade.

Summary

• The GPU is a "time machine": By significantly accelerating computation through parallel processing, it enables scientists to complete research that would originally take a lifetime within their own lifetime, essentially allowing them to "see the future."
• The success of CUDA comes from "evidence-based hope": Believing that GPUs, due to their vast gaming market, will become the world's most produced parallel processors, thereby providing powerful computing resources for more researchers.
• The next decade of AI is the decade of "applied science": With foundational model capabilities already evident, the focus will shift to applying AI in specific fields such as digital biology, climate science, agriculture, and robotics to solve real-world problems.
• Anything that can move will be automated: Through the combination of Omniverse (a virtual world simulation platform) and Cosmos (a world foundational model), robots will efficiently learn in the digital world, accelerating their entry into the physical world.
• Everyone needs an "AI mentor": AI will lower the barriers to knowledge, empowering individuals to leverage the capabilities of "super AI" to focus on more creative and valuable work, becoming "super humans."

From Gaming Graphics Cards to Computing Revolution: The Birth of the GPU and the "Time Machine" Philosophy

At the beginning of the conversation, Cleo Abram pulled the timeline back to the 1990s. At that time, game developers were eager to create more realistic graphics, but the hardware could not handle the massive mathematical calculations. NVIDIA's foundational insight was that in software programs, only about 10% of the code was responsible for 99% of the processing work, and this 99% of processing could be executed entirely in parallel. The remaining 90% of the code needed to be executed sequentially. Jensen Huang pointed out that an ideal computer should be able to handle both sequential and parallel processing, rather than choosing one or the other. This observation became the starting point for NVIDIA: to build a company that solves problems ordinary computers cannot address.

To intuitively illustrate the difference between CPU (central processing unit) and GPU (graphics processing unit), Jensen Huang mentioned a video on NVIDIA's official channel that has been around for 15 years. In it, the "MythBusters" team demonstrated with two robots: a small robot firing one paintball at a time, representing the sequential processing of the CPU; while another large robot fired all the paintballs simultaneously, representing the parallel processing of the GPU. This parallel processing capability first released tremendous energy in the video game industry.

Why choose gaming as a breakthrough? Jensen Huang provided three reasons: first, the team loved this application, and the simulation of virtual worlds is appealing; second, they foresaw that video games had the potential to become the largest entertainment market in history (which has proven true); third, a vast market means considerable R&D budgets to support ongoing innovation in complex technologies. The flywheel created between technology and market ultimately propelled NVIDIA to become one of the most important tech companies in the world. "Everything started with video games," Jensen Huang summarized.

More profoundly, Jensen Huang likened the GPU to a "time machine." He shared a story: a quantum chemistry scientist told him, "Jensen Huang, because of NVIDIA's work, I can complete my lifetime research within my lifetime." This is time travel—compressing work that would exceed a lifetime into one life. Whether it's weather forecasting, virtual city traffic simulation, or self-driving car testing, seeing the future through simulation is the "time travel" brought by parallel computing. The GPU not only makes computations faster but also enables people to "see" the future in advance.

The Belief in CUDA: Building Bridges, Awaiting AI's "Explosion"

After achieving success in parallel processing in gaming, researchers began to try utilizing the powerful computing capabilities of GPUs in other fields, such as CT reconstruction in medical imaging. However, in the early days, researchers had to "trick" the GPU into treating scientific computing problems as graphics problems. This process was very challenging. The internal driving force at NVIDIA stemmed from the gaming world itself: to create more dynamic and realistic virtual worlds (such as flowing water, realistic explosions), they needed to process particle physics, fluid dynamics, and more, which far exceeded the capabilities of traditional graphics pipelines.

Multiple factors converged to give birth to CUDA. This is a platform that allows programmers to use familiar languages like C to directly command GPUs for general computing. Jensen Huang explained that the core belief in creating CUDA was: due to the vast scale of the gaming market, NVIDIA's GPUs would become the most produced parallel processors in the world. Therefore, CUDA, built on this architecture, had a significant opportunity to reach the most people. This is a strategy based on "evidence-based hope": if you don’t build it, they (developers, researchers) will never come.

The turning point came in 2012. Ilya Sutskever, Alex Krizhevsky, and Geoff Hinton from the University of Toronto trained AlexNet using a CUDA-based GeForce GTX 580 GPU and achieved an overwhelming victory in the renowned ImageNet image recognition competition. AlexNet, a type of neural network, marked the shift in the computing paradigm from "step-by-step instruction computation" to "training computers to learn from massive examples."

Jensen Huang recalled that NVIDIA was also working on solving computer vision problems at the time but was facing setbacks. The emergence of AlexNet surprised and excited them. They asked themselves: if AlexNet could achieve this in computer vision, what are its limits? They inferred that if deep learning architectures could be scaled, then the vast majority of machine learning problems could be represented using deep neural networks. This could potentially reshape the entire computing industry. It was this insight that prompted NVIDIA to redesign the entire computing stack, leading to the launch of DGX AI supercomputers. "Sixty-five years after the IBM System 360 introduced modern general computing, we reinvented what we know as computers," Jensen Huang said.

The following decade was one of steadfast belief. Jensen Huang admitted that there were difficult moments during this time, with huge investments and uncertainty from the outside world. But based on the two core beliefs of "accelerated computing" and "deep learning scalability," they persisted in investing billions of dollars. "Why did we persevere for so long? The answer is quite the opposite: we had no reason not to persevere because we believed in it." This belief has driven NVIDIA through cycles until the AI wave fully erupted.

Physical AI and Robotics: Training "Everything that Moves" in the Digital Universe

Looking to the future, Jensen Huang divided the development of AI into two phases: the past decade was "the foundational science of AI," while the next decade will be "the applied science of AI." AI will be applied in countless specific fields such as digital biology, climate technology, agriculture, fisheries, logistics, and education.

One key application direction is "physical AI," which means robotics. Jensen Huang predicted: "Everything that moves will one day be a robot, and it will happen soon." The key to driving this transformation lies in a fundamental change in training methods. In the past, training robots was either conducted in the fragile real world or relied on limited data sources like motion capture. Now, by training in the digital world, robots can undergo repetitive exercises far beyond physical world possibilities in just one day, with exponential improvements in learning speed.

NVIDIA has created the Omniverse (a platform for 3D world simulation and collaboration) and Cosmos (a world foundational model) for this purpose. Jensen Huang used ChatGPT as an analogy to explain this system: although the early versions of ChatGPT were powerful, they could produce "hallucinations" (generating content not based on facts). Later versions could obtain "ground truth" by uploading PDFs or searching online, generating more reliable answers.

Similarly, for robots to understand the physical world, they need a world foundational model that comprehends concepts such as gravity, friction, inertia, geometric space, object permanence, and causality—this is Cosmos. Omniverse provides a precise simulator based on Newtonian physics principles, serving as Cosmos's "ground truth." The combination of the two allows the system to generate an infinite number of physically plausible future scenarios for robots to learn from. "Just as PDF plus ChatGPT can answer countless questions, Omniverse plus Cosmos can do the same for the physical world."

This means that robots in factories can instantaneously simulate all possible paths and unexpected situations (darkness, congestion, etc.) in a digital twin environment without any wear on actual equipment. Jensen Huang envisions a future where everyone has their own "R2-D2," which could take the form of smart glasses, smartphones, cars, or physical robots at home, always by their side, growing together.

Empowering "Super Humans": AI Mentors and Core Suggestions for the Future

In the face of such rapid technological change, how should ordinary people prepare? Jensen Huang offered clear and positive advice.

First, he encouraged everyone to immediately find an "AI mentor." This AI mentor can teach you anything, help you program, write, analyze, think, and reason. AI significantly lowers the knowledge barrier for understanding any field, allowing everyone to have a "personal mentor at their fingertips." Jensen Huang used his own example, stating that being surrounded by thousands of "super humans" (top experts) in his company has never made him feel unnecessary; rather, it has given him the confidence to tackle grander goals. Similarly, being surrounded by "super AI" skilled in specific tasks will empower and enhance everyone's capabilities. "We will become super humans, not because we have superpowers, but because we have super AI."

Next, he suggested that every student and professional should ask themselves a central question: "How can I use AI to do my job better?" Jensen Huang recalled that his generation was the first to have to ask, "How can I use computers to do my job better?" And for the next generation, this question will be transformed into, "How can I use AI to do my job better?" This question is universally applicable, whether for lawyers, doctors, chemists, or biologists; it's both a starting point and an endpoint for thought.

He particularly emphasized the importance of "prompt engineering." Interacting with AI, akin to being skilled in asking questions, is a discipline that requires both skills and artistry. Learning how to effectively prompt AI to become your capable assistant is a critical skill.

Lastly, Jensen Huang discussed the fundamental limitation of energy efficiency. He showcased a small AI supercomputer, pointing out that since the first DGX-1, valued at $250,000 and delivered to OpenAI in 2016, performance has increased sixfold while energy efficiency has astonishingly improved by ten thousand times. Pursuing higher computation efficiency is the primary task in creating smarter systems.

As the conversation drew to a close, when asked how he hoped people would evaluate his work, Jensen Huang's response was concise yet powerful: "They made an extraordinary impact." He realized that NVIDIA, due to its long-standing core belief, has grown to become one of the most important and influential tech companies in the world. They take this responsibility seriously and strive to ensure that the capabilities they create, regardless of profitability, company size, or fame, benefit large businesses, independent researchers, developers, and every scientific field. He hopes that in the future, when people look back, they will find that digital biology and materials science have been transformed because of NVIDIA, that robots are everywhere assisting humans, and that the epicenter of all this is precisely that company whose games they grew up playing.

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。