Will AI Replace Humans? History Suggests Power Matters More

Overview

Anthropic’s recent economic scenarios for 2030 suggest a world where US GDP rises by a third while office work plummets, yet manual trades like plumbing see wage growth. This divergence highlights a critical assumption: the models do not account for rapid advancements in robotics, potentially underestimating the displacement of blue-collar labor.

However, historical patterns of technological fear often obscure the real drivers of economic change. While predictions of imminent automation have frequently failed, structural forces such as market elasticity and political power dynamics ultimately determine who benefits from new technologies.

Robots and Cassandras

In October 2015, prominent tech leaders made bold predictions about the timeline of autonomous vehicles. Sam Altman and Elon Musk claimed self-driving cars would fill streets within two to four years, while Ray Kurzweil had forecasted the same for 2019. Conversely, taxi drivers remained skeptical, arguing that AI lacked the dexterity to handle complex physical tasks like changing tires.

History suggests these early AI prophets were overly optimistic about near-term physical automation. By 2026, robot taxis remain confined to specific geographic pockets and weather conditions, validating the skepticism of workers in traditional driving roles.

The failure of general-purpose robots like Baxter further illustrates this gap. Designed to undercut low-skilled labor in manufacturing, the robot struggled with basic assembly tasks and was discontinued in 2018. Its limited success demonstrates that physical labor remains significantly harder to automate than intellectual work.

The Jevons Paradox in White-Collar Work

While robotics lagged, advancements in narrow AI have disrupted professional services. In 2016, Geoffrey Hinton predicted that AI would replace radiologists within five years by outperforming humans in reading scans. Ten years later, the US has more radiologists than ever, and their wages have increased.

This phenomenon is explained by the Jevons paradox, first observed in 1865 regarding steam engine efficiency. When technology makes a resource more efficient, consumption of that resource often increases rather than decreases. Similar patterns have appeared across various sectors:

• Radiology: Automated scan reading led to a surge in diagnostic testing, creating higher demand for radiologists to manage the increased volume and patient care.
• Legal: Automating document review expanded the scale of litigation and discovery, increasing the need for paralegals.
• Banking: ATMs reduced the cost of cash transactions, but banks responded by opening more branches and shifting tellers toward complex customer service and sales.

Anthropic’s own reports emphasize automating specific tasks rather than entire professions, suggesting that work will evolve rather than vanish. However, this historical comfort may not apply to the current wave of large language models.

The Uniqueness of Generative AI

Unlike previous automation tools, large language models claim a degree of general intelligence that could disrupt a broader range of cognitive labor. While narrow AI replaced specific steps in processes, LLMs can handle diverse, unstructured tasks that were previously considered safe from automation.

The pace of adoption is already evident in emerging technologies. Waymo has clocked 500,000 paid rides weekly in San Francisco, capturing a significant local market share. Similarly, Baidu’s Apollo Go operates over 300,000 rides weekly across China and the Middle East. Although these figures represent less than one percent of national vehicle miles traveled, they have grown exponentially in just two years.

Meanwhile, humanoid robotics is making a resurgence in both the US and China. While the industrial cobot era saw setbacks, new iterations are entering factories and consumer markets, suggesting that the barrier between intellectual and physical automation is beginning to dissolve.

The primary distinction in this cycle is the scope of generalization. Previous technologies augmented specific skills; current AI threatens to automate the cognitive glue that holds many white-collar professions together.

Basic Income and the Balance of Power

Proponents of technological optimism often suggest that displaced workers will transition to creative pursuits or rely on universal basic income. However, the historical implementation of social welfare suggests that such benefits are granted out of political necessity rather than altruism.

Otto von Bismarck introduced health insurance and pensions in Germany during the 1880s to counter the rising influence of socialist movements and secure worker loyalty to the state. Likewise, Henry Ford doubled wages in 1914 to reduce the 370% annual turnover rate caused by the monotony of the assembly line.

Social progress, therefore, is often tied to the leverage held by the workforce. When workers are indispensable, they can demand better conditions and compensation. When they become surplus, political systems may resist redistribution unless stability is threatened.

Hoplites and Rowers: A Lesson in Power

The dynamics of power and representation can be traced back to classical Athens. Before 462 BC, political influence was limited to hoplites—citizens who could afford expensive armor. The introduction of the trireme fleet democratized power, as naval strength depended on rowers from the lower classes rather than wealthy landowners.

Power tends to follow utility. As the military and economic importance of a specific class shifts, so too does their political influence and the social concessions they receive.

Athens’s democracy expanded when rowers were essential to the state’s survival, only for power to revert to the hoplite elite after naval defeats. This cyclical nature of progress indicates that rights and protections are not permanent gifts from above but are earned through the strategic value of labor.

Conclusion

The question of whether AI will replace humans cannot be answered solely by examining technological capability. While history shows that markets often absorb automation through increased demand and new roles, the current generation of general AI poses a broader challenge than previous narrow tools.

Ultimately, the fate of workers depends less on the speed of AI development and more on the balance of power. Just as ancient rowers gained political voice because their labor was vital to the state, modern workers will negotiate their position based on their indispensability in an increasingly automated economy.