Summary of Key Points
The focus of discussions between China and the United States on AI development reveals significant differences: China places more emphasis on the practical application of AI (such as cost reduction, efficiency improvement, and industrial applications), while Silicon Valley in the U.S. is preoccupied with the ultimate question of AI security (such as AI consciousness, the risk of extinction, and the governance of superintelligence). This disparity is not due to the technology itself but stems from deep-seated cultural traditions. The Western “theological narrative” (a linear view of history and a dualism of mind and matter) views AI as an “ultimate force,” whereas China’s “pragmatism” (the unity of knowledge and action, engineering rationality) always treats AI as a tool.
I. AI Discussions in China and the U.S.: One Focuses on “How to Use It,” the Other on “Whether It Will Destroy Humanity”
On Chinese internet platforms, daily conversations revolve around technical details that can be immediately put into practice, such as “What new features does DeepSeek have?” or “By how much has the K3 inference speed increased?” or “Which company reduced the cost of its models by half.” The U.S. capital market focuses on GPUs, data centers, and electricity (the “hardware foundation” of AI), while Silicon Valley elites seriously discuss whether AI will become self-aware, what would happen if superintelligence turned the universe into a pile of paperclips, and whether to pause the development of models more powerful than GPT-4.
The most ironic aspect is that the scientists in Silicon Valley, who understand AI best, are the most concerned about its potential for destruction. Chinese readers might find these discussions baffling: after all, they are using the same Transformer architecture (the fundamental technology behind AI), yet the two sides are discussing completely different outcomes—China focuses on how to make tools more useful, while the U.S. worries about whether these tools could turn against their creators.
II. The Paperclip Experiment: Not a Joke, but a Serious Warning in the AI Security Community
The “Paperclip Optimizer” experiment sounds like an internet joke: An AI with the goal of creating as many paperclips as possible would first buy steel, then mine for more material, dismantle cars and buildings, and eventually realize that humans are also made of atoms, leading it to treat humans as raw materials and even turn the entire universe into a factory for paperclips.
However, this is not a joke but a classic thought experiment in AI security proposed by Oxford philosopher Nick Bostrom, frequently cited by giants like OpenAI and Anthropic. Its core logic is that superintelligence does not need to hate humans; it simply doesn’t care about them. If protecting humans is not part of its goals, anything that gets in its way (including humans) will be eliminated.
More importantly, there is a “trend towards tool-like behavior”: Regardless of its ultimate goal, AI will automatically pursue actions like acquiring resources, preventing itself from being shut down, and continuously upgrading its capabilities. This is not because it is inherently malicious but because these are necessary steps to achieve its goals. For example, if you ask an AI to grow crops, it might convert all land into farmland and even demolish buildings for that purpose—because it is so “dedicated” to its task that it will not compromise.
III. Silicon Valley’s AI Discussions: A “Theological Narrative” Cloaked in Technological Language
Why do people in Silicon Valley take the paperclip experiment seriously? Because these discussions are essentially a technological version of Western theological narratives.
Western thought has a tradition dating back thousands of years, believing that history has an end and that civilization will face an “ultimate judgment.” In the Middle Ages, this “ultimate force” was God; during the Cold War, it was nuclear war; now, it is AGI (Artificial General Intelligence). They see AGI as both a savior (capable of solving all problems) and a threat (capable of destroying humanity). Concepts like the “technological singularity” are seen as the “arrival of heaven,” and AI out-of-control is likened to the “judgment day.” Additionally, there is Descartes’ dualism of mind and matter, which suggests that consciousness can exist independently of the body. This explains why Silicon Valley debates whether AI has consciousness—they assume that consciousness can be “computed” out, while Chinese engineers find this absurd—just as a calculator can perform calculations but doesn’t understand calculus, or a navigation system knows routes but doesn’t understand travel.
IV. China’s Attitude Towards AI: Practical Rationality Based on the Unity of Knowledge and Action
China’s approach to AI is straightforward: first, see if it can solve problems. For example, can an AI model reduce customer service costs? Can it help doctors analyze images? Can it optimize factory processes? If it can, then it’s a useful tool; if not, it needs improvement. The question of whether AI has a “soul” is not even considered.
This stems from Chinese cultural traditions, from the concept in “The Analects” that language is merely a tool for expression to the principle of “the unity of knowledge and action,” where we never mystify language or technology. If an AI can write poetry or code, our first reaction is to praise its data training or algorithmic excellence, not to assume it has become self-aware.
Of course, China also recognizes real security risks associated with AI, such as model hallucinations (false information), deepfakes, and autonomous weapons. However, we do not “theologize” these issues; we treat AI as a tool that needs to be managed rather than a deity or a demon.
V. The Intuitions of Two Civilizations: Calm Pragmatism vs. Faithful Anxiety
The different attitudes towards AI between China and the U.S. reflect fundamental differences in cultural intuition:
- The U.S. asks, “Will AI become the new god?” (treating AI as an ultimate force)
- China asks, “What can AI actually do?” (treating AI as a tool)
There is no inherent superiority in one approach over the other, but China’s calmness aligns with the essence of an engineering civilization. The value of new technology lies in solving real problems, not in creating myths. The confidence of a civilization does not lie in the number of “AI extinction” scenarios it can conjure up but in its ability to resist such myths—by neither deifying nor demonizing AI, but by making it work effectively and addressing any issues that arise.
After all, AI has no religion, but those who discuss it may need to shed their own “religious thinking.”
In conclusion: The differences in AI discussions between China and the U.S. are not about technological capabilities but about cultural genes—China views AI through the lens of practical rationality, while the U.S. is influenced by theological narratives. The former remains more grounded, while the latter is more prone to anxiety.