Summary of Key Points
This article compares the economic philosophies of Keynes (demand-side management) with those of Amodi in the era of AI (efficiency-driven growth), highlighting how the AI revolution is rapidly replacing knowledge-based workers and exacerbating wealth inequality, leading to a resurgence in the crisis of "insufficient effective demand." It argues that the AI era requires new Keynesian redistribution measures—such as an AI tax, Universal Basic Income (UBI), and reduced working hours—to redistribute the wealth created by AI among the general population, thereby maintaining economic stability and avoiding the pitfalls of the Great Depression.
I. Why Did Keynes Implement "Digging Holes to Fill Them Again"? — The Lessons of the Great Depression: Technology Moved Too Fast, Distribution Didn't Keep Up
In the 1920s, a technological boom in the United States (with the widespread adoption of electricity and internal combustion engines) doubled factory production, but workers' wages did not increase, and wealth concentrated in the hands of a minority (1% owning 40% of the wealth). As a result, factories were filled with goods that ordinary people could not afford to buy, breaking the economic cycle—this is what Keynes referred to as "insufficient effective demand." Keynes's approach of "digging holes to fill them again" was not arbitrary; it involved the government spending money to create jobs (even if they were unproductive) to provide income for the general population and restart the economic cycle. This strategy helped the middle class grow and the economy stabilize over the following 30 years. However, this approach was later replaced by laissez-faire policies, leading to renewed wealth concentration until the problems of the AI era emerged.
II. This Time, AI Is Different: Machines Are Taking Over Mental Work, and the Middle Class Is at Risk
The Industrial Revolution replaced manual labor (such as that of farmers and workers), but AI is targeting cognitive tasks—jobs held by white-collar professionals such as programmers, accountants, legal assistants, and designers. For example, Claude Code allows one engineer to perform the work of a team; on GitHub, 70% of every 100 lines of code are written by AI. Companies like IBM plan to cut 30,000 back-office jobs, and Goldman Sachs reports a monthly net loss of 16,000 positions (more jobs being replaced than created). The problem is that AI-driven layoffs are abrupt, while new job creation is slow, leaving ordinary people without income and thus devoid of purchasing power.
III. AI Companies Are Making Huge Profits, but the Money Isn't Going to the General Population — Why Does Keynes Matter Again?
Companies like Anthropic generate annual revenues of $44 billion with profit margins exceeding 70%, but where does all that money go? Most of it goes to company shareholders and a few employees, not to the workers whose jobs have been replaced. The wealth created by AI is concentrated in the hands of a minority (10% of American households own 70% of the stock market). Without consumer demand from the general population, who will buy the products produced by AI?
This is precisely what Keynes fought against: the failure of algorithms to distribute wealth fairly. No matter how efficient a system is, if most people lack purchasing power, all businesses will fail. Therefore, Keynesian logic is essential—government intervention is needed to redistribute the excess profits generated by AI among the general population.
IV. New Approaches to "Digging Holes to Fill Them Again" in the AI Era: How to Distribute the Benefits of AI
The article proposes several specific measures to distribute the benefits of AI:
1. AI Excess Profit Tax: The government should levy a higher tax on the excess profits generated by AI companies that exceed industry averages.
2. UBI Pilots: In regions where AI is widely used (such as California), implement an AI tax to distribute cash monthly to residents, not as charity but to provide them with a basic level of consumption power.
3. Reduced Working Hours: If AI can perform 40 hours of work, reduce the standard working week to 32 hours to create more jobs and share the benefits of productivity.
4. Public AI Funds: The government should invest in AI companies using public resources (such as computing power and data) and distribute the dividends to those affected by AI, making everyone a "shareholder" in these technologies.
These measures are not about "robbing the rich to help the poor" but about maintaining system stability. Without consumer demand from the general population, even the most advanced AI will be unable to sell its products, leading to a collapse for everyone.
Conclusion
AI represents the future, but without Keynesian mechanisms to ensure social equity, it could lead to economic collapse. We need to apply Keynesian redistribution principles to harness the efficiency of AI and ensure that the benefits are shared by the majority. Otherwise, the wealth created by AI will become a digital game for a few, ultimately dragging down the entire economy.