Written by: Techub News Compilation

Introduction

Recently, NVIDIA founder and CEO Jensen Huang appeared again on the interview program "Special Competitive Studies Project." Within less than a year, Huang returned to the same stage and engaged in an in-depth discussion with the host about the significant changes that have taken place in the AI field over the past few months. As a core driver of the global AI computing wave, Huang's remarks not only interpret the evolution of AI technology but also provide highly forward-looking industry insights from multiple dimensions, including national strategy, industrial policy, and technological security.

Summary

• AI is a five-layer "cake" composed of energy, chips, infrastructure, models, and applications, and the United States should strive to maintain global leadership at each layer.
• AI is a once-in-a-lifetime market force driving the "reindustrialization" of America, creating trillions of dollars in manufacturing output and high-skilled job opportunities.
• Physical AI (such as robots and autonomous driving) is just around the corner, with breakthroughs relying not only on AI models but also on advancements in materials science and hardware like motors.
• Regarding chip export restrictions to China, Huang believes policies need to keep pace with the times, and completely giving up the Chinese market lacks strategic significance, while the U.S. still maintains a lead at the model layer.
• AI will not eliminate jobs but will change the nature of "tasks" in work; the current "fear narrative" surrounding AI may hinder technological adoption, which is detrimental to the U.S.

From Retrieval-based Computing to Generative Computing

Huang began by pointing out that we are in a completely new computing paradigm. Traditional computing is essentially "retrieval-based": people pre-create content (text, images, videos), store it, and then retrieve it through search engines, recommendation systems, etc. Whether it is shopping recommendations or video streaming, the core logic is matching and recall.

The fundamental change brought about by artificial intelligence is "generative computing." Systems dynamically generate entirely new content that did not exist before based on the user's context, requests, and intentions, starting from an initial information seed (possibly based on certain facts or reports). Each generated result may differ significantly. The benefit of this paradigm shift is that computing can provide the most relevant information to meet user needs based on real-time situations and changes in the real world.

This breakthrough requires a fundamental shift in computing architecture: from the need for large amounts of storage to the need for massive computing power. At the same time, AI systems need to learn to perceive various types of information (text, video, images), understand user intentions and contexts, and plan how to address questions or complete tasks. This marks our transition from an era of "retrieving information" to an era of "generating intelligence."

The Five Layers of the "Cake" of AI and America's Reindustrialization Opportunity

Huang used the metaphor of a "five-layer cake" to deconstruct the AI industry. From bottom to top, they are: energy, chips, infrastructure, models, and applications.

The bottom layer, energy, is crucial because the enormous AI computing consumes vast amounts of electricity. Huang used this to introduce a broader topic: America's reindustrialization. He pointed out that American society and the economy face a critical choice: should they revitalize manufacturing so that people who do not require a four-year college degree can also gain high-value job opportunities? He believes AI is an unprecedented market force driving this historic change.

AI will spur the construction of three types of factories: first, chip manufacturing and packaging factories; second, computer system assembly factories; and finally, "AI factories" that deploy these computers. This will bring trillions of dollars in manufacturing output and a large number of high-skilled, high-paying job positions. The essence of manufacturing is to change the form of materials, which requires a tremendous amount of energy. Therefore, if the U.S. wants to seize the opportunity of reindustrialization, it must simultaneously upgrade its energy network, improve grid efficiency, and develop sustainable backup energy.

Huang emphasized that the U.S. was a leader in technology application during past industrial revolutions and must not become a laggard in this industrial revolution centered around AI. He specifically pointed out that within the five layers of the cake, the "application and adoption" layer is most critical for the U.S. It is essential to ensure that the U.S. remains at the forefront of applying this technology, as it directly relates to productivity, prosperity, and economic leadership.

Breakthroughs in Agent AI and Physical AI

When discussing specific advancements in AI technology, Huang believes that the key breakthrough from large language models (LLMs) to chatbots is "reinforcement learning from human feedback" (RLHF). The leap from LLMs to agent systems lies in the ability to "harness" these models.

The latest agent systems can connect to the real world (ground truth), conduct web searches, research, reason, possess memory, and communicate with other agents. Over the past six months, agents have made significant progress. Huang particularly praised the outstanding performance of coding agents like Codium and Claude Code, believing that most software tasks can now be automated.

This raises a common concern: will AI replace software engineers? Huang’s answer is no, and he draws an analogy with radiologists. A decade ago, predictions suggested that computer vision would replace radiologists; today, AI has indeed permeated all aspects of radiology, yet there is a heightened demand for radiologists. The reason is that "reading scan images" is a "task" in a radiologist's work, while their "purpose" is to diagnose diseases. AI has taken over some tasks but has enhanced the doctors' ability and value in achieving their core purpose.

Similarly, a software engineer's "task" is programming, but their "purpose" is innovation, problem-solving, and identifying needs. AI has automated coding tasks, allowing engineers to focus more on high-value innovative activities, thereby increasing the demand for more software engineers. Huang warned that spreading the fear that "AI will eliminate jobs" is dangerous and destructive, as it could drive away the talent that society urgently needs.

Regarding the next wave—physical AI, Huang expressed optimism. He believes the "scientific issues" for autonomous taxis (Robo-taxis) have been solved; the main issue now is engineering implementation, and NVIDIA's "Drive" platform and "Thinking Car" technology are driving it toward reality.

Regarding humanoid robots, Huang provided a vivid analogy: if today's video generation models can create a video of "picking up a coffee cup and taking a sip" on command, then logically, generating commands for a robot to perform the same actions should also be possible. The primary challenges have partially shifted to mechatronics, including motors, mechanical hands, materials (which need to be lightweight yet sturdy), battery technology, and sensors. He hinted that this technology is "just around the corner."

Chip Exports to China, Open Source, and Security

Being in Washington, Huang inevitably addressed the sensitive issue of chip export controls to China. He reiterated the "five-layer cake" framework, arguing that America's goal should be to become a global leader at every layer and to actively export technology to the world, thus expanding the global influence of the American tech stack.

In response to the argument "to stop China's AI development by limiting computational power," Huang stated that every layer is important; while the chip layer is indeed critical, completely surrendering a massive market like China "lacks strategic significance." He believes related policies need to be dynamically adjusted to keep pace with the times. At this stage, allowing American chip companies to operate in the Chinese market is reasonable.

He acknowledged that China is already a global leader in energy (production and related technologies), is closely following the U.S. at the AI model layer, and has a vast number of AI researchers, which is its "national treasure." The U.S. must continue to attract global talent and maintain focus on this.

Huang's biggest concern lies in the AI application and adoption layer. He criticized American discourse on AI as being overly "sci-fi" and "alarmist," creating unnecessary fear, while other regions in Asia are enthusiastically embracing and adopting AI. He warned that this could lead the U.S. to lag behind in application.

Regarding the safety of open-source AI, Huang holds a positive attitude. He believes that open source can actually enhance security and safeguards. In the future, resisting super AI network attacks may require relying on large-scale defensive agent groups trained on open-source models, using asymmetrical advantages to compete. Open source allows technology to develop in the sunlight, facilitating review and ensuring safety.

Regarding the risks that open-source projects like "OpenClaw" might pose, NVIDIA has developed the "OpenShell" technology to provide a secure "sandbox" environment for agents, controlling their information access and output, and integrating privacy protection strategies. This solution has been contributed to the open-source community and adopted by numerous companies, aiming to allow enterprises to deploy AI agents more safely.

Eliminating Fear and Embracing Opportunities Brought by AI

At the end of the interview, Huang strongly urged society to view AI more rationally and positively. He refuted extreme statements that "AI poses an existential threat to humanity" or "will destroy a large number of jobs," calling such claims "absurd" and "counterproductive."

He cited facts: AI has created over 500,000 jobs in recent years; it is the best opportunity for America's reindustrialization and reshaping of manufacturing; companies that adopt AI grow faster, leading to more hires. The core of the problem lies in the confusion between "tasks" and "purposes" in work.

He humorously used himself as an example: if only the "task" is considered, his job is "clicking on a phone and talking," which AI has long been able to do, so he should have been unemployed by now. But the reality is that he is busier than ever. Because his "purpose" is leading the company, formulating strategies, and driving innovation, which AI cannot replace.

Huang concluded that AI liberates humans from tedious "typing" tasks, allowing us to focus more on exercising our imagination and solving larger-scale problems. He posed the rhetorical question: if we can act faster, aspire higher, and have greater expectations through the empowerment of AI, how could this be bad for the nation? This is exactly the state we strive for. He called on all sectors to actively adopt AI, as this technology has become not only "amazing" but also genuinely "useful," and will continue to amplify human potential and ambition.

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。