Bullshit Jobs and Smart Machines

This post explores how many of today’s high‑paid professions depend on collecting and analysing data, and on decisions made on the basis of that process. Drawing on thinkers such as Hannah Arendt, Gerd Gigerenzer, and others, I examine the paradoxes of complex versus simple algorithms, the ethical dilemmas arising from algorithmic decision‑making, and how automation threatens not only unskilled but increasingly highly skilled work. I also situate these issues in historical context, from the Fordist assembly line to modern AI’s reach into law and medicine.

Originally published in Substack: https://substack.com/inbox/post/170023572

Many contemporary highly paid professions rely on data gathering, its analysis, and decisions based on that process. According to Hannah Arendt (2017 [original 1963]), such a threat already existed in the 1950s when she wrote:

“The explosive population growth of today has coincided frighteningly with technological progress that makes vast segments of the population unnecessary—indeed superfluous as a workforce—due to automation.”

In the words of David Ferrucci, the leader of Watson’s Jeopardy! team, the next phase in AI’s development will evaluate data and causality in parallel. The way data is currently used will change significantly when algorithms can construct data‑based hypotheses, theories and mental models answering the question “why?”

The paradox of complexity: simple versus black‑box algorithms

Paradoxically, one of the biggest problems with complex algorithms such as Watson and Google Flu Trends is their very complexity. Gerd Gigerenzer (2022) argues that simple, transparent algorithms often outperform complex ones. He criticises secret machine‑learning “black‑box” systems that search vast proprietary datasets for hidden correlations without understanding the physical or psychological principles of the world. Such systems can make bizarre errors—mistaking correlation for causation, for instance between Swiss chocolate consumption and number of Nobel Prize winners, or between drowning deaths in American pools and the number of films starring Nicolas Cage. A stronger correlation exists between the age of Miss America and rates of murder: when Miss America is aged twenty or younger, murders committed by hot steam or weapons are fewer. Gigerenzer advocates for open, simple algorithms; for example, the 1981 model The Keys to the White House, developed by historian Allan Lichtman and geophysicist Vladimir Keilis‑Borok, which has correctly predicted every US presidential election since 1984, with the single exception of the result in the Al Gore vs. George W. Bush contest.

Examples where individuals have received long prison sentences illustrate how secret, proprietary algorithms such as COMPAS (“Correctional Offender Management Profiling for Alternative Sanctions”) produce risk assessments that can label defendants as high‑risk recidivists. Such black‑box systems, which may determine citizens’ liberty, pose enormous risks to individual freedom. Similar hidden algorithms are used in credit scoring and insurance. Citizens are unknowingly categorised and subject to prejudices that constrain their opportunities in society.

The industrial revolution, automation, and the meaning of work

Even if transformative technologies like Watson may fail to deliver on all the bold promises made by IBM’s marketing, algorithms are steadily doing tasks once carried out by humans. Just as industrial machines displaced heavy manual labour and beasts of burden—especially in agriculture—today’s algorithms are increasingly supplanting cognitive roles.

Since the Great Depression of the 1930s, warnings have circulated that automation would render millions unemployed. British economist John Maynard Keynes (1883–1946) coined the term “technological unemployment” to describe this risk. As David Graeber (2018) notes, automation did indeed trigger mass unemployment. Political forces on both the right and left share a deep belief that paid employment is essential for moral citizenship; they agree that unemployment in wealthy countries should never exceed around 8 percent. Graeber nonetheless argues that the Great Depression produced a collapse in real need for work—and much contemporary work is “bullshit jobs”. If 37–40 percent of jobs are such meaningless roles, more than 50–60 percent of the population are effectively unemployed.

Karl Marx warned of industrial alienation, where people are uprooted from their villages and placed into factories or mines to do simple, repetitive work requiring no skill, knowledge or training, and easily replaceable. Global corporations have shifted assembly lines and mines to places where workers have few rights, as seen in electronics assembly in Chinese factory towns, garment workshops in Bangladesh, and mineral extraction by enslaved children—all under appalling conditions.

Henry Ford’s Western egalitarian idea of the assembly line—that all workers are equal—became a system where anybody can be replaced. In Charles Chaplin’s 1936 film Modern Times, inspired by his encounter in 1931 with Mahatma Gandhi, he highlighted our dependence on machines. Gandhi argued that Britain had enslaved Indians through its machines; he sought non‑violent resistance and self‑sufficiency to show that Indians did not need British machines or Britain itself.

From industrial jobs to algorithmic threat to professional work

At its origin in Ford’s factory in 1913, the T‑model moved through 45 fixed stations and was completed in 93 minutes, borrowing the idea from Chicago slaughterhouses where carcasses moved past stationary cutters. Though just 8 percent of the American workforce was engaged in manufacturing by the 1940s, automation created jobs in transport, repair, and administration—though these often required only low-skilled labour.

Today, AI algorithms threaten not only blue‑collar but also white‑collar roles. Professions requiring long training—lawyers and doctors, for example—are now at risk. AI systems can assess precedent for legal cases more accurately than humans. While such systems promise reliability, they also bring profound ethical risks. Human judges are fallible: one Israeli study suggested that judges issue harsher sentences before lunch than after—but that finding has been contested due to case‑severity ordering. Yet such results are still invoked to support AI’s superiority.

Summary

This blog post has considered how our economy is increasingly structured around data collection, analysis, and decision‑making by both complex and simple algorithms. It has explored the paradox that simple, transparent systems can outperform opaque ones, and highlighted the grave risks posed by black‑box algorithms in criminal justice and financial systems. Tracing the legacy from Fordist automation to modern AI, I have outlined the existential threats posed to human work and purpose—not only for low‑skilled labour but for highly skilled professions. The text argues that while automation may deliver productivity, it also risks alienation, injustice, and meaninglessness unless we critically examine the design, application, and social framing of these systems.

---

References

Arendt, H. (2017). The Human Condition (Original work published 1963). University of Chicago Press.
Ferrucci, D. (n.d.). [Various works on IBM Watson]. IBM Research.
Gigerenzer, G. (2022). How to Stay Smart in a Smart World: Why Human Intelligence Still Beats Algorithms. MIT Press.
Graeber, D. (2018). Bullshit Jobs: A Theory. Simon & Schuster.
Keynes, J. M. (1930). Economic Possibilities for our Grandchildren. Macmillan.
Lee, C. J. (2018). The misinterpretation of the Israeli parole study. Nature Human Behaviour, 2(5), 303–304.
Lichtman, A., & Keilis-Borok, V. (1981). The Keys to the White House. Rowman & Littlefield.