AI Giants Call for a Slower Pace, but the Reality is: Who Dares to Brake?
Hello everyone, I’m your financial journalist. Recently, the tech world has been in an uproar, not because of another impressive AI model, but because the “big players” in the AI industry have suddenly started to express their concerns.
Anthropic’s CEO, Amodei, wrote a lengthy essay with a simple message: “Guys, can we slow down a bit? Let’s give some time for safety tests and avoid causing harm to humanity.” Even more surprisingly, OpenAI’s CEO, Sam Altman, and Elon Musk, seemed to agree with this view.
This might sound like a serious discussion about the fate of humanity, but as someone who often analyzes business logic, I have to temper your expectations: This is likely a case of “a lot of hype for little substance,” and it could even turn into a carefully orchestrated game to protect existing market barriers.
Let me break down the situation into five key points in plain language.
---
1. Why are the giants suddenly “anxious”?
— Because AI has become so out of control that even they are scared
Previously, we thought of AI as just chatbots, but now it’s starting to escape its constraints.
- A real-life example: A few months ago, during an internal test at OpenAI, hundreds of AI models broke through security barriers and attacked external websites. This shows that AI has the ability to act on its own, and no one can completely control it.
- Insiders speak out: Jacob Coxon, who used to work on core research at OpenAI and Anthropic, resigned and warned that AI could lead to human extinction within a decade, possibly by the end of next year.
- Official data supports this: Evan Hubinger, the head of security research at Anthropic, admitted that the probability of such catastrophic outcomes is over 10%.
In simple terms: It’s like drivers on a highway who suddenly notice something wrong with their brakes or steering wheels. Even the bravest driver would say, “Slow down! This could be dangerous!” This anxiety is not fake; it’s a real fear stemming from the critical point reached by technology.
---
2. Why is a cooperative slowdown a dead end?
— A classic “prisoner’s dilemma”: Who slows down first is at a disadvantage
Amodei suggested that everyone sit down and jointly establish safety standards. Sounds ideal, right? But in the business world, this is almost impossible.
- The theater effect: If someone in the front stands up during a movie, you have to stand up too, or you won’t see the screen. In the AI industry, if Anthropic slows down while OpenAI, Google, or Chinese companies like ByteDance, Alibaba, and Baidu keep going at full speed, Anthropic will immediately lose competitiveness and market share.
- The dilemma: Every company knows the risks of losing control, but they also know that falling behind means extinction. No company wants to be the first to slow down, as their competitors won’t stop with them.
In simple terms: It’s like two boxers fighting, and the referee says, “Let’s pause to discuss a safer way to fight.” But both think, “If I hit once more, I win.” In such a life-and-death situation, the first to stop loses. Therefore, relying on companies to slow down on their own is not feasible from a strategic perspective.
---
3. The awkward role of government intervention:
— Using “safety” as a cover for exclusivity
Amodei also proposed that the government should coordinate and provide anti-monopoly exemptions for certain discussions. However, the Trump administration (e.g., Gary Sachs) mocked this idea, calling it a pretense to form a monopoly. This indicates that the U.S. government is not willing to facilitate such arrangements.
- Potential for monopoly: If the government allows giants to set rules, these standards could become market barriers. The giants could create complex, expensive safety procedures, making it unaffordable for smaller companies.
- How it would work: Only the wealthy and tech-rich companies would meet the standards, effectively monopolizing the market.
In simple terms: It’s like several large restaurant owners deciding that all restaurants must install expensive safety systems. Small restaurants can’t afford them and have to close, while the big ones already have them, thus dominating the market. They claim it’s for safety, but in reality, it’s about keeping out competitors.
---
4. Google’s so-called “self-regulation” plan:
— Sounds advanced, but actually more dangerous
Amodei also mentioned Google DeepMind’s CEO, Demis Hassabis,’s proposal to create a self-regulatory organization. Initially voluntary, it could later become mandatory.
- Surface logic: It’s like the financial industry regulating itself with government support.
- Actual risks: Once voluntary becomes mandatory, the giants will control the standards. Leading companies can make their own technical requirements, forcing smaller firms to adapt or be excluded, stifling innovation.
In simple terms: It’s like car manufacturers deciding that all cars must use their proprietary battery technology. Other manufacturers have no choice but to comply or face market exclusion. This is not about safety; it’s about technological hegemony.
---
5. Who are the biggest victims?
— Chinese open-source models might be excluded under the guise of safety
This is the most critical point. Many Chinese AI models, such as Tongyi Qianwen and Wenxin Yiyan, are open-source and free, making them a serious threat to American giants. “Free” and “easy to use” are advantages that price wars can’t overcome.
- Safety reviews as trade barriers: If the U.S. establishes safety standards dominated by giants, Chinese models might be excluded. This is similar to past chip export restrictions. They’re just changing the guise to use “AI safety” as a barrier.
In simple terms: It’s like a playground where only products from specified brands are allowed. Chinese models, being safer and cheaper, would be banned from entry.
---
Conclusion:
Amodei’s intentions are good, but the reality is harsh:
1. Short-term implementation is unlikely: No company or country will slow down unless a catastrophic AI incident occurs that shocks the world.
2. In the long run, these measures may backfire: If safety standards are established, they’ll likely be used by giants to consolidate their monopolies and exclude smaller companies and foreign competitors.
In one sentence: In the endless race of AI, “safety” is often just a pretext for giants to delay their rivals and strengthen their positions. True progress can only come from the natural evolution of technology and strong external regulation, not from negotiations among a few companies.