Hello, I'm your financial news analysis assistant. This article about Jensen Huang's rebuttal of the "AI doomsday theory" at the All-In summit is incredibly informative. It doesn't just discuss the debate over technological paths but also delves deeper into conflicts of business logic, cultural differences, and regulatory philosophies.
To help you easily understand this "verbal battle" among the top figures in Silicon Valley, I will first summarize the key points and then break down the content in five dimensions.
---
【Summary of Key Points】
In one sentence:
Jensen Huang, CEO of NVIDIA, publicly refuted the recently popular "AI doomsday theory" (the idea that AI could lead to human extinction or the end of civilization) at the All-In summit. While he doesn't deny the risks associated with AI, he strongly opposes using "precise probabilities" (such as a 10% extinction rate) to quantify these unknown risks, criticizing such claims for lacking empirical evidence from engineering and even suggesting they could hinder technological progress.
Core arguments:
1. Opposition to a binary choice: Security and development are not opposing concepts; the United States can maintain its lead in AI while also ensuring proper safety measures.
2. Criticism of pseudoscientific predictions: Predictions from the past decade that AI would replace doctors, programmers, etc., have mostly proved incorrect. The current "extinction probabilities" are similarly unfounded and constitute irresponsible scaremongering.
3. Engineering mindset vs. philosophical anxiety: Huang advocates using engineering methods (such as third-party audits and power regulation) to address specific issues, rather than getting caught up in metaphysical fears.
4. Interests and positions: He suggests that some AI companies (like Anthropic) may be hyping up risks on the eve of their IPOs, potentially engaging in a form of "regulatory capture"—using public fear to create barriers that only giants can overcome.
5. Cultural differences: The West, influenced by cultures of original sin and doomsday judgments, tends to fear technology; in contrast, Huang, with his engineering background, prefers a problem-solving and pragmatic approach.
---
【In-depth Analysis: Five Dimensions for a Layman**
1. "Predictions are free, but implementation costs money": Two completely different views of risk
This is the key to understanding Huang's stance. We need to distinguish between predictors and engineers:
- Predictors (such as some AI security researchers):
They sit in offices, logically concluding that "AI has a 10% chance of leading to human extinction within the next 10 years."
- Cost: Almost zero. If they're wrong, no one loses money; if they're right, no one gets a bonus.
- Consequences: If society believes them, it may stop investing and restrict research and development. The cost of this "brake" falls on the entire industry and society.
- Huang's analogy: This is what he calls the "externalities of prediction." For example, when Turing Award winner Stuart Russell Turing predicted that radiologists would be replaced by AI in 2016, 10 years later, there was a severe shortage of radiologists in the U.S., making it harder for patients to get treated. The prediction was wrong, but the cost was borne by ordinary people.
- Engineers (like Huang):
Their judgments must be translated into real-world outcomes. For instance, they need to predict how many units of the next generation of chips (Rubin) will be sold, how much HBM memory to purchase, and how much power to allocate.
- Cost: Extremely high. A wrong prediction could lead to NVIDIA's bankruptcy.
- Mindset: Huang often says, "The company could go bankrupt in 30 days." This kind of fear is concrete, immediate, and solvable. It forces them to deliver products and fix bugs today.
- Conclusion: Huang dislikes the "doomsday theory" because it's abstract, distant, and impossible to eliminate through effort. It encourages inaction, while an engineer's fear drives action.
2. The lesson from radiologists: Why we should be wary of "precise nonsense?"
Huang cited a series of examples to show that seemingly confident and precise AI predictions are often the least reliable:
- Case studies:
- 2016: Stuart Russell predicted the end of radiology.
- 2023: Some claimed GPT-2 and Llama 3 were too dangerous to release.
- Recently: Claims that half of low-level jobs would disappear within 6-9 months, or that 90% of code would be generated by AI.
- Real outcomes:
Radiologists haven't disappeared; instead, the profession is in high demand due to a shortage of trained professionals.
- Deep logic: These predictors are often well-educated researchers, but "understanding theory" doesn't equate to "understanding reality." The real world is full of complexities like friction, policies, human behavior, and infrastructure constraints. Those who give 10% extinction probabilities simplify complex systems engineering into mathematical problems. Huang believes this "over-certainty" is dangerous as it misleads policymakers and the public and can be used to hinder competition.
3. The suspicion of "regulatory capture": Who is using "safety" as a guise for monopoly?
This part of the article is highly insightful from a political and business perspective. Although Huang didn't directly name Anthropic, his words suggest that the current "AI safety" discourse might be a carefully crafted business strategy:
- Coincidental timing: Anthropic (owned by Dario Amodei) just submitted its IPO application, valued at $965 billion, and is expected to go public soon.
- Amodei published an article advocating slowing down the pace of cutting-edge AI and calling for stronger regulation.
- What is regulatory capture? This is an economic concept where large companies use their resources to push for strict regulations that only they can meet.
- Logical chain: If regulations require top-tier computing power or expensive third-party audits, smaller companies, startups, and open-source communities will be excluded.
- Result: The market will be dominated by a few giants (like OpenAI, Anthropic, Google, Meta). They act as both athletes and referees, raising barriers to protect their profits and monopolies.
- Huang's counterargument: He supports regulation but based on facts. He believes the real risks come from a few labs with tens of thousands of GPUs. Regulation should target these "big players" rather than creating universal, cumbersome barriers.
- He proposes financial audits: Independent third parties should review AI companies' code, data, and processes, just like auditors review financial reports. This ensures safety without giving AI companies control over their own standards.
4. Cultural differences: The Prometheus myth vs. the Chinese approach to technology
The article explores cultural differences that explain why Americans (especially Silicon Valley elites) have a quasi-religious fear of AI, while Huang remains calm and pragmatic:
- Western narrative (original sin and judgment):
- Mythological examples: Prometheus stealing fire, Frankenstein's monster, the forbidden fruit in the Bible.
- Core logic: Knowledge/technology is a sin that will lead to judgment or destruction.
- Consequences: This mindset leads to avoidance, restriction, and repentance. It leads to extreme statements like "airstriking data centers" because there's an underlying belief that such technology shouldn't exist.
- Effective altruism (EA): Some scholars view the future as a mathematical expectation and believe that any risk, even if small, must be prevented at all costs.
- Chinese/Engineer narrative (tools and problem-solving):
- Mythological examples: The story of Pangu creating fire, Yu the Great controlling floods, and the concept of "man-made technology."
- Core logic: Technology is a neutral tool; its use determines its morality. Instead of waiting for disaster, we should control it.
- Huang's attitude: "Since it will be created, how can we use it correctly?" This reflects a sense of responsibility rather than doomsday fear.
- Huang's personal background: His mother taught him to memorize through repetition, and his father was a chemical engineer, emphasizing practicality.
5. The true AI competition: Not about who is stronger, but who uses it best
Finally, Huang redefines the meaning of the AI competition and outlines his outlook for the future:
- Open source vs. closed source: He emphasizes the importance of open-source models. In the past six months, $400 billion in venture capital flowed into AI companies, 80% of which relied on open-source models.
- If only closed-source models were used, only giants would win. If open-source models are strong (many contributed by China), everyone, including teachers and small companies, can participate.
- American strategy: To win, the U.S. needs to involve the whole society; open source is the foundation of inclusive technology.
- What's the real competition? Not about who has the largest model or reaches AGI first, but about who uses technology best.
- Historical analogy: Basic sciences like electromagnetism and electricity originated in Europe, but the U.S. turned them into industrial revolution engines through engineering and infrastructure.
- In the AI era: Basic models can come from anywhere (including China), but the winner will be those who integrate AI into data centers, power grids, cars, and factories.
- De-mystifying superintelligence: Huang believes we may already be in the AGI era (human-level intelligence). Superintelligence doesn't require an all-knowing entity; it's already present in specific applications like autonomous driving and protein design.
- Final attitude: "It's our responsibility to build it well. If the risks are real, solve them through engineering; if they're not, don't scare people. Don't stop progress due to unverified fears.**
---
**Implications for You**
1. Don't be scared by precise probabilities: When someone tells you AI has a 10% chance of destroying the world, think of the radiologist example. Such figures often aim to create anxiety, not provide solutions.
2. Focus on implementation, not concepts: For most people, whether AI destroys the world is irrelevant; what matters is how it affects your job. Engineering (creating and fixing things) is more important than theory (predicting the future).
3. Be skeptical of safety claims with financial motives: If a company calls for regulation while preparing for a IPO, question whether those regulations are only affordable for giants.
4. Embrace open source and participation: Don't think of AI as a game for big companies. Open-source models make technology more accessible to everyone. True competitiveness lies in how you use it, not in how much computing power you have.
5. Maintain an engineer's mindset: In the face of uncertainty, don't fall into doomsday anxiety. Ask, "What can I do to solve this problem?" This mindset is more powerful in both work and life.