Summary of the Core Content
This article discusses the concept of the “permanently marginalized class”: The development of AI may strip most people of their bargaining power in the workplace. Those without capital will not only become poorer but also lose the means to move up socio-economically—such as by earning enough to buy assets or negotiating higher wages through unions. This could lead to a situation where they remain “permanently at the bottom.” When asked about the future of an average 17-year-old, AI researchers are unable to provide a clear answer because individual efforts (such as learning AI or purchasing AI-related equity) do not address the underlying issues. What is truly needed are new distribution systems and negotiation mechanisms, which currently do not exist in the United States. While China has made some attempts, the key will be whether it can establish such frameworks in time.
Detailed Analysis
1. “Permanently Marginalized Class”: Not Just Poverty, but the Loss of Ladders to Ascend
In the past, those at the bottom of society were poor, but they still had opportunities to improve their situation: they could earn wages through labor, save money for education or housing, and gradually move up. Workers could also negotiate with employers to share in the benefits of increased productivity. However, AI is different. If AI can perform most cognitive and physical tasks, employers will need fewer workers, reducing the proportion of wages in the economy and shifting wealth towards those who own AI models, computing power, and data centers. As a result, ordinary people will lose these opportunities: they won’t be able to earn enough to buy assets through labor, nor will they have the leverage to negotiate profit sharing with employers (since employers no longer depend on their work). Stanford data shows that employment among 22-25-year-olds in AI-related fields has decreased by 16%, and the unemployment rate for recent graduates has risen by 1.6 percentage points—it’s as if the “floor buttons” of elevators have been removed, leaving no way to move upward.
2. Researchers Are Stumped by Questions About the Future
When asked how an average 17-year-old should prepare for the future, AI researchers respond with “we don’t know.” This is not due to their ignorance but a recognition of the limitations of individual actions:
- Learning AI: The more skilled you become in using AI, the less valuable your labor becomes, potentially accelerating the elimination of other workers.
- Purchasing AI equity: Ordinary people cannot afford shares in companies like OpenAI; only employees and wealthy individuals can buy them. Even if they do, future AI systems might render these assets worthless (for example, what would a future super-AI or the state recognize your past investments for?).
The real issue is the lack of social rules that ensure those not exploited by AI can share in its benefits.
3. Fear of Replacement Leads to a Vicious Cycle
Those aware of the risks of AI-driven job loss continue to work in AI-related fields, and many scholars and policymakers switch to AI research labs. They do this to secure a position within the tech industry (by acquiring equity). However, this creates a cycle: more people fear losing their jobs, leading them to invest in equity, further weakening labor’s bargaining power. For example, AI PhDs from Berkeley choose easier research topics to get hired faster, and policymakers leave their positions to work for AI companies, leaving workers with even less protection.
4. Individual “Escape Routes” Are Illusory
Common solutions, such as learning AI, buying AI equity, or pursuing physical jobs, have drawbacks:
- Learning AI: You may help companies become more efficient, but this leads to job cuts and a faster decline in overall labor value.
- Buying AI equity: Ordinary people cannot afford it, and future ownership rights might be invalid.
- Physical Jobs: These roles may still be automated (e.g., with AI-assisted medical diagnoses).
5. The Solution Lies in New Negotiation Mechanisms
To address the issue of the “permanently marginalized class,” we need systemic changes:
- Laws: For example, Chinese courts have ruled that AI cannot directly replace jobs, and state-owned companies are required not to use AI for layoffs.
- Unions: New types of unions should be established to involve workers in negotiations about AI-related job transitions and compensation.
- Taxation/Profit Distribution: AI companies should be taxed or required to share profits with the public through taxes or funds.
The current lack of such mechanisms in the U.S. has led to extreme reactions (such as people throwing bottles and shooting at legislators). China has made some small-scale attempts. The next two years will determine whether the U.S. can establish these frameworks before tensions escalate.
In conclusion, what that 17-year-old really needs is an explanation of why their fate depends on whether they are valued by capital. However, researchers cannot provide this answer because it requires systemic changes, not individual actions. For now, people can only try to protect themselves by acquiring equity in AI-related companies.