Summary of Key Points
Recently, there has been a major incident that has gone viral in the global AI community, with millions of views: Over the past four days, seven core researchers from three of the world's top AI institutions—OpenAI, Anthropic, and Google DeepMind (half of whom are responsible for AI safety and governance)—have publicly stated that the super AI systems that the industry is desperately working on to develop could pose a threat to human survival by 2030, or even lead to human extinction. Coxon, a former researcher at Anthropic who took the lead in making these statements, voluntarily gave up a substantial portion of his company's stock, which he would have received in two months, essentially sacrificing millions in profits to speak the truth and clearing himself of any suspicion of manipulating the company's stock price for traffic generation. This revelation doesn't just involve ordinary industry gossip; it exposes a long-hidden dilemma within the AI community: the very people who understand AI best are simultaneously terrified that their creations could destroy humanity and yet cannot stop their research and development efforts. Behind this lies a significant problem with the industry's incentive mechanisms, which also serves as a wake-up call for everyone: we don't need to oscillate between the extremes of "AI deification" and "AI apocalypse"; it's time for us to demand a say in the development of AI.
---
Detailed Analysis
1. This AI warning is nothing like the usual marketing ploy
In the past, when AI companies claimed their technology was dangerous, people often saw it as a form of boast: "Look how powerful our technology is that it could destroy the world—hurry up and buy our services or invest in our stocks." However, this time is different:
- Coxon, the person who revealed the information, voluntarily gave up his stock, showing no motive to profit from inflating the company's value;
- Among the seven researchers who spoke out, there were those in charge of scalable supervision and alignment science at Anthropic, essentially the people hired to prevent AI from getting out of control. It's like core engineers who built the atomic bomb coming out and saying, "The thing we're creating could really blow up the Earth"—this is not the nonsense of laypeople;
- The old joke in Silicon Valley that "AI professionals leave to write poetry and spend time with their families" was once used to mock the high pressure and extreme risks of the industry, but now it's becoming a reality, indicating that the panic within the community can no longer be hidden.
2. The absurd paradox of the industry: Those afraid of AI extinction are actually accelerating its development
Many people are confused by this: If these people know AI could be deadly, why don't they simply stop working on it? It's not hypocrisy; the entire industry has fallen into an inescapable "accelerated death cycle":
The logic of all leading AI companies is the same: if I slow down or stop, those unscrupulous companies or bad actors will develop super AI first, making the situation even more dangerous. Therefore, I must stay ahead to ensure safety. As a result, everyone uses the pretext of "leading to protect humanity" as a moral justification for their rapid progress, and the industry's speed only increases, with no one daring to brake. In this system, the companies reap all the benefits, while the consequences of potential disasters fall on all of humanity. Expecting industry professionals to act out of conscience is unrealistic.
3. Don't be led by the "apocalyptic narrative": The immediate harms of AI are more real than the potential extinction in 10 years
After reading the news, some people are too scared to sleep, while others think the researchers are just trying to attract attention. However, there's a more pressing issue: the widespread discussion about whether AI will extinguish humanity in 10 years obscures a more immediate and significant problem:
Everyone focuses on the distant threat of extinction, ignoring the harms that AI is already causing to individuals. AI scams have led to many people losing everything, AI-powered interviews are rejecting countless low-income job seekers, and a few leading AI companies hold more power over users and decision-making than small nations. These issues are happening right now, without waiting for AI to become self-aware and cause direct harm. Moreover, if the public focuses only on the distant endgame, it will create a situation where only the companies that develop AI can define its dangers, leaving the public with no say and further empowering those companies.
4. As ordinary people, we don't need to take sides; we just need three clear rights
Public opinion often pushes people into two extremes: either believing the researchers are just trying to boost stock prices or assuming AI is about to destroy the world and giving up. There's no need for that. A survey of over 2,700 top AI researchers globally shows that 38% believe the probability of AI leading to human extinction is at least 10%, and 70% think that investment in AI safety research is insufficient. This is not the delusion of a few madmen, but there's no concrete evidence that AI extinction is inevitable. We don't need to understand complex concepts like large model parameters or alignment technology; we just need to demand three things:
- When AI companies develop powerful new models, they should not launch them without independent third-party safety audits unrelated to their interests;
- Employees who discover safety risks within the companies should not be silenced by confidentiality agreements; there must be safe, anonymous channels for reporting;
- If AI causes problems, the companies should bear full responsibility, not ordinary people.
After all, no one has the right to gamble with the fate of humanity for their own business success, even if they claim to be acting for humanity's sake.