This book review is also available in French, in Revue internationale du Travail 165 (2), and Spanish, in Revista Internacional del Trabajo 145 (2).

The Eye of the Master by Matteo Pasquinelli has a central thesis: artificial intelligence (AI) does not emulate biological intelligence, but rather, the intelligence of labour and social relations.

According to Pasquinelli, computer science theories are not neutral, nor have they been developed in the abstract. On the contrary, they have inherited a subtext of colonial fantasy and class division from the industrial era. As an example, he cites Turing, who divided computing tasks into masters, servants and girls.¹ AI seeks to capture knowledge expressed through individual and collective behaviours and encode it in algorithms in order to automate a wide range of tasks. Ultimately, the most sophisticated intelligent machines are based on imitating the division of labour.

Charles Babbage plays a central role in Pasquinelli’s thesis. In the early nineteenth century, Babbage sought to replace the repetitive work of “computers” (usually women who performed various mathematical calculations from their homes and sent them by post) with an automated steam-powered machine. This device, called a “difference engine”, automated logarithmic calculations and produced error-free logarithmic tables, which were crucial for astronomy and the hegemony of the British Empire. In short, the aim was to mechanize “mental labour”. The two principles that guided Babbage’s work were: (1) the labour theory of the machine – i.e. machines replace a previous division of labour – and (2) the principle of labour calculation, or Babbage’s principle, which postulates that the division of labour allows the exact amount of work required to be calculated and purchased. These ideas of Babbage’s were already present in the work of Adam Smith, who described the division of labour with his famous example of a pin factory and suggested that many machines were invented by imitating the organization of tasks in the workplace (Smith [1776] 2012). According to Babbage, calculating machines are also tools for measuring, disciplining and monitoring work.

Another author relevant to Pasquinelli’s thesis is Karl Marx, who questioned technodeterminism by arguing that the main drivers of capitalism were workers and the division of labour (Marx [1867] 2013).

Machines imitate and reproduce human labour, which also has an impact on scientific progress, according to Pasquinelli (pp. 90–91), who states categorically:

Tool-makers and machine operators knew that they were contributing to the invention of new technologies. What they were rarely aware of is that they were also contributing to new scientific discoveries. New machines prompt scientific notions and paradigm shifts more often than science happens to invent new technologies from above. As in an example mentioned earlier, it was the steam engine which gave birth to thermodynamics, rather than the other way around.

Pasquinelli argues that artificial neural networks do not imitate the physiology of the brain, as is often suggested, but rather, self–organize information. Indeed, cybernetics was born out of the idea of information as a means of feedback with the environment and internal self-regulation. This perspective was based on Jakob von Uexküll’s early 20th-century view of biology, which saw the organism as an information-processing system struggling to adapt to the environment. In fact, the word “cybernetics”, coined by Norbert Wiener, referred to the ability of a technical, social and living system to control itself through the exchange of information with the environment.

Pasquinelli emphasizes that cyberneticists projected the technical composition of their era – comprising telegraph networks, electromagnetic relays, feedback systems and television scanners – onto their conceptualizations of the brain and nature. They likewise transposed forms of self-organization that were already present in the division of labour and the technical organization of their societies onto nature.

In other words, information technologies do not reshape society, as is often claimed, but rather, social relations forge communication networks, information technologies and cybernetic networks from within.

Pasquinelli highlights the importance of Friedrich Hayek’s thesis on connectionism and its influence on the paradigm of artificial neural networks. Then, in the final chapter, he analyses the evolution of Frank Rosenblatt’s “Perceptron”, a statistical neural network for pattern recognition, emphasizing its influence on deep learning. The author discusses the rivalries between the symbolic AI paradigm and the connectionist AI of pattern recognition, which has ultimately prevailed. Rosenblatt’s “Perceptron” is also part of a broader genealogy of work automation, pioneering the automation of a different type of work: perception or supervision work. It should come as no surprise that this perception and supervision are functional to discipline and authority in the workplace. According to Pasquinelli (p. 246):

Ultimately, AI is not only a tool for automating labour but also for imposing standards of mechanical intelligence that propagate, more or less invisibly, social hierarchies of knowledge and skill. As with any previous form of automation, AI does not simply replace workers but displaces and restructures them into a new social order.

Pasquinelli’s thesis is thought-provoking. It is unsurprising that the division of labour has been a source of inspiration for machines, automation and AI, although all these advances have evolved in parallel with efforts to improve weapons of war. Pasquinelli mentions the latter in passing, although it should be considered a factor as significant as the division of labour in the development of technologies.

Pasquinelli’s work is counter–narrative because it challenges the dominant view of AI and because it connects with many of the concerns about its impact. We could distinguish three competing narratives put forward by science fiction literature: that of Frankenstein, that of Čapek and that of Asimov.

The dominant narrative is that of Mary Shelley’s Frankenstein, first published in 1818. Supposedly, the creators of AI, like Dr Frankenstein, have played at being gods by attempting to imitate general human intelligence. Their creation, AI, has rebelled and wreaked havoc by replacing jobs, discriminating through algorithms and facilitating “surveillance capitalism” (Zuboff 2019).

Karel Čapek’s narrative in his 1920 play RUR (Rossum’s Universal Robots) is radically different. Čapek was the first to use the term “robot”, which in Czech means “slave, servant”, and in his play, robots are created exclusively to work and replace workers, although they eventually rebel. Pasquinelli explains in his book that the twist in the AI plot consists precisely of an idea that has endured since Babbage’s time: imitating and better controlling the organization of work, disciplining workers and then replacing them.

An intermediate narrative is that of Isaac Asimov, who explained his three laws of robotics as an attempt to counteract the pessimistic view of Shelley and, above all, Čapek. Robots are machines that can be used for good or evil, and the laws of robotics proposed by Asimov sought to demonstrate that they could be designed to not cause harm (1990).

How should we face the future? One way would be to consider Asimov’s proposal in a less utopian light. As Pasquinelli suggests in his conclusions, political steps must be taken to ensure that the design and development of AI take into account community and collective interests.

How does Pasquinelli challenge us as labour law academics? His thesis addresses us from several perspectives. First, we must leave anthropomorphism behind. We must not succumb to the mirage and slogan of “singularity”. More than 170 years ago, Ada Lovelace was already dismissing anthropomorphism, and today the propaganda of singularity seems like a marketing strategy for technology companies to gain more wealth and power (Cortina 2024). AI is a technology that mimics human intelligence, but is far removed from any notion of singularity (Cortina 2024). Therefore, the ultimate responsibility for its use must lie with someone – that is, a human being (Bertolini 2022).

Second, greater trust should be placed in legal mechanisms. AI does not operate on its own; it has been created within a social context, with a specific intention (conscious or unconscious), based on the division of labour. There is a significant difference between using an intelligent system and delegating decisions that are crucial to people’s lives to such a system. (Cortina 2024).

The rhetoric of an imminent AI “singularity” and paradigm shift encourages the belief that everything must change and that established structures, such as traditional labour law, are no longer valid. Every era requires the law to adapt, but this does not mean that one should feel powerless and anxious in the face of these phenomena. Just as fear can wreak havoc on democracies, the rhetoric of paradigm shift – and the dismissal of legal mechanisms as obsolete – can hinder the search for the dynamism needed to respond effectively to change.

The gig economy provides a good example of this. When platform work first emerged, it seemed to signal the rise of a new type of worker operating outside institutional and legal frameworks. However, the principle of the primacy of reality (Gamonal and Rosado 2019) has provided a good starting point for addressing the issue of driver classification from within an existing legal framework (Kocher 2022). I am not suggesting that legal categories should not evolve, but that apocalyptic discourses should be set aside, and the incredible advances in AI should be assessed on their own terms. Pasquinelli helps to remove the magical veil from AI and to promote regulatory proposals for AI anchored in traditional labour law mechanisms, as De Stefano does for collective bargaining, social dialogue and algorithms (2019).

Discussions about new technologies and their impact on the world of work have been a recurring theme in the history of labour law. Therefore, this issue should not be approached from a blank slate. On the contrary, it is necessary to move away from apocalyptic discourse and to adapt existing legal mechanisms to the new reality, drawing on the debates and ideas developed in earlier periods of labour law (Cherry 2019).

Third, the discussion should not focus solely on technical aspects, but also on the political sphere. We tend to forget the importance of power (Gamonal 2024). Pasquinelli emphasizes that AI is political and that we must develop “an AI policy” that promotes the common good. Several authors have argued that technologies should be designed to serve human needs (Acemoglu and Johnson [2023] 2024; Albin 2024). Moreover, in the twentieth century, there are examples of alternatives that challenged technological determinism in the workplace (Honneth 2024). However, for Pasquinelli, this is not just an ideal aimed at redirecting, but a necessity to correct a course that has been planned since the time of Babbage. We must follow the advice of Pasquinelli and Asimov and leave behind the technological determinism that makes us powerless spectators of its advances and ravages.

Notes

1. In the early nineteenth century in England, the term “computer” was used to refer to administrative employees, who were generally women, who did tedious mathematical calculations by hand (p. 51). ⮭

References