Summary of Key Points
This article primarily discusses the concern that AI could lead to widespread and permanent unemployment (the "AI employment apocalypse"). However, as AI is implemented in practice, technology leaders, economic theories, and empirical data have all shown that AI will not eliminate a large number of jobs; rather, it will change the nature of work. Nevertheless, there are challenges to consider, such as the speed at which society can adapt to AI, the experience required for entry-level positions, and wealth distribution, which are more pressing issues.
1. Technology Leaders Converge on the View: AI is Not the Culprit of Unemployment, but a Tool for Work Transformation
Three technology industry leaders (the CEOs of Goldman Sachs, NVIDIA, and OpenAI) have refuted the "apocalypse" theory from different perspectives:
- Goldman Sachs CEO Solomon: AI could reduce working hours by 25% in the next decade, but it will not eliminate 25% of jobs. Similar to how the widespread use of Excel and Zoom has made tasks more complex rather than less, AI will replace repetitive tasks in accounting and banking, making the work itself more sophisticated (for example, banks using AI to streamline account opening processes may hire more staff to interact directly with customers).
- NVIDIA CEO Huang Renxun: He criticized bosses who use AI as an excuse for layoffs, pointing out that while AI has only become practical in the past six months, many companies have been laying off employees for two years, so the timing does not match. He argued that the new jobs created by AI are roughly equivalent to those replaced, and AI is not yet capable of causing widespread unemployment.
- OpenAI CEO Altman: A former strong supporter of the "apocalypse" theory, he now believes that entry-level white-collar positions will not be eliminated because human interactions (such as empathy in customer communication) cannot be replicated by AI, so there will be no employment crisis.
2. Economics Debunks the "Apocalypse" Theory: The Total Amount of Work is Not Fixed; New Needs Will Create New Jobs
The "apocalypse" theory makes a classic mistake—the labor supply fallacy: assuming that the total amount of work in the world is fixed and that machines will leave people without jobs. In reality:
- Demand Increases: As technology reduces costs, new demands emerge. For example, the invention of cars reduced the number of horse-drawn carriages but created more jobs for drivers, mechanics, and gas station employees. Similarly, AI makes programming faster, leading to more programming projects and thus a greater need for programmers.
- Flexible Demand Comes to the Rescue: Some jobs (like programming and legal research) are subject to flexible demand; when costs decrease, people are more willing to engage in these activities. For instance, AI helps lawyers quickly find information, allowing them to take on more cases without losing jobs. Fixed-demand jobs (such as payroll processing) are more likely to be replaced by AI, but they represent a smaller proportion.
- Keynes' Prediction: The father of macroeconomics, Keynes, predicted that automation might reduce working hours to 15 per week, but people would use the extra time to pursue new careers (such as live streaming and AI training).
3. Data Supports the View: AI is More of a "Assistant," and the Employment Market Remains Stable
Empirical data refute the "apocalypse" theory:
- Companies Use AI as an Assistant: A U.S. media analysis of corporate financial reports found that mentions of AI as an enhancement tool were eight times more frequent than those referring to it as a replacement tool, indicating companies prefer using AI to improve efficiency rather than lay off employees.
- Stable Employment Trends: Since 2025, software development jobs have been growing, and product management positions have rebounded to their highest levels from 2022. The number of new businesses and newly launched applications is also increasing. The National Economic Research Council in the U.S. states that AI has not significantly impacted employment, with 90% of companies reporting no effect on jobs.
- High Cost of AI: Uber's CTO noted that the company's budget for AI tools was exhausted within four months, and a NVIDIA vice president mentioned that current AI computing costs are higher than labor costs, so companies are cautious about using AI to replace workers.
4. Challenges Cannot Be Ignored: The Pain Caused by AI Depends on Society's Ability to Adapt
Refuting the "apocalypse" theory does not mean there are no challenges. The real issues include:
- Adaptation Speed: Can society quickly develop new skills and adjust its employment systems to accommodate AI? For example, with fewer entry-level jobs, how can young people gain experience (for instance, in accounting, where new employees learn from existing ones)?
- Wealth Distribution: Who will benefit from the money created by AI? If company profits increase but employee wages do not, it could exacerbate wealth inequality.
- Risk of Economic Recession: Historically, many jobs have disappeared during economic downturns, not due to technology, but as a result of economic decline. If AI replaces some jobs during a recession, it could lead to structural unemployment.
What are the solutions? David George suggests redesigning apprenticeship programs for the AI era (e.g., teaching skills for human-computer collaboration) and involving companies, schools, and governments in helping workers transition.
In Conclusion
AI will not cause humans to become unemployed; instead, it will radically change the way we work. The key is whether we can quickly adapt and turn AI into a tool for our benefit rather than a threat. Instead of worrying about whether we will lose jobs, we should focus on practical issues such as how to distribute resources, train new workers, and prepare for economic downturns.