After de-skilling, art?
One response to full automation
There is clear evidence that AI is taking human jobs: entry level, technical, creative, meaningless, possibly even management. It will affect every job, according to various human resources experts. I say this as a potential victim myself, having worked across some of the most vulnerable areas — art, design, teaching, coding, research.
Can it be stopped? I’m pessimistic. But one possible response comes from an unlikely source: the art world more than a century ago. In this article I will explain how art might be an answer to automation, by zooming in from the big economic picture to what we might do individually and collectively.
Gaining productivity, losing meaning
I was talking with Saleh Kayyali, who always has a good view of commercial developments in tech. At the individual level, he says, AI can shift our role from active participant to passive monitor. You know this if you do any coding, commercial art or design these days.
Amazon workers might stand still for 12 hours instead of whizzing around the warehouse, with robots lining up to serve them tasks — but with adverse effects on mental health. When pilots operate drones, or surgeons use robotic arms, their work becomes safer, but boring and alienating, according to research. In all cases, work can feel meaningless — tasks don’t necessarily become more creative, just more repetitive. You need to work faster, and might have more anxiety about job security.
In order to find some useful big-picture insights, Kayyali said, we need to navigate between the marketing hype of AI companies and the doomsayers. Academics are generally warning of impending doom. AI companies are too, while at the same time promising endless growth. Both sectors are, however, informing national policy in various countries. AI companies selling things that may be either useful or harmful is one thing, but when governments go all in without giving us a say, this touches all citizens’ lives, and livelihoods.
Some countries seem to have an economy based on growth at all costs, and there has always been a disconnect between the corporate and national drive for continuous growth on one hand, job security and worker wellbeing on the other. (I wrote a little primer on techno-capitalism here.)
Social media is full of examples of how AI can enable people to do new things, and to make complex tasks much easier, so there is a case for more jobs and more productivity. But does it mean bringing people into jobs that might eventually be fully automated away? On the other hand, the companies tell us that over the longer term, no one will need to work at all in a world where AI has also eliminated hunger, poverty, disease, global warming, etc. Let’s dig deeper into some research.
Tools, rules, repetition
Matteo Pasquinelli has an insightful analysis, breaking this down into labor, rules, and automation. He reverses the Marxist view, arguing that social relations drive technical means of production, not the other way around. He shows that labor already became mechanical and logical well before automation. Tools from factory machines to AI not only automate labour but impose “standards of mechanical intelligence [his italics] that propagate, more or less invisibly, [already existing] social hierarchies of knowledge and skill”.
Given that different cultures can have very different social relations, Kayyali adds, we are likely to see different forms of AI and automation in different countries.
Zooming in to the organisational level, Pasquinelli’s division of automation, rules and labor maps neatly onto Yrjö Engeström’s use of tools, rules and the division of labour to analyse collaboration. Engeström’s focus is on tool mediation, context of the activity, and contradictions that might arise from these relations.
To illustrate this, take software development. In a corporate context oriented to shipping a product intended to serve customers effectively, AI as a tool mediates what coders, project managers, and marketing teams do — specifically by automating processes intended to result in increased productivity. But conflicts arise from contradictions between, for example, tools and the division of labour: Kayyali sees, for example, the roles of designers and coders merging. Vibe coders versus highly trained (and highly paid) software engineers. Or tools versus rules, as when AI makes increased productivity the new normal, or eliminates roles altogether.
Productivity itself can be seen as a contradiction between tools, rules and community. AI boosts efficiency, and efficiency alone is not enough to increase productivity. It might reduce the time you spend on email, but it doesn’t necessarily increase creative or collaborative work. Economist Carl Benedikt Frey observes that, as computers have gotten faster, economic growth has actually slowed. If AI enables us to do more stuff, we invest less time in each thing, but new innovations tend to come from a single-minded focus on one narrow domain, not from doing things faster and spreading yourself thinner.
Even AI helping us to do that one thing faster won’t help. Since AI is trained on past data, Frey points out, “A model trained before Galileo would have parroted a geocentric universe; fed 19th-century texts it would have proved human flight impossible before the Wright brothers succeeded.” An AI-filled world still needs humans to invent new things (not to mention maintaining all those AI systems).
Noam Scheiber observes that industrial automation didn’t simply eliminate jobs, but broke them down into repeated, mechanical tasks, as mentioned above. So the things that humans are still better at doing than machines also became mechanised: skilled workers gave way to assembly lines, individual secretaries became typing pools.
Scheiber argues that the same is happening with AI in software development, reporting that corporate coders’ work is “becoming more routine, less thoughtful and, crucially, much faster paced.” Productivity goes up, companies push coders to use AI, and this has “in effect, eliminated much of the time the developer spends reflecting on his or her work.”
Work goes into the toilet
John Roberts, in his 2010 article “Art After Deskilling,” analyses how industrialization in the mid- to late-1800s not only gave artists new things to depict in their work, but transformed how they made it, and how it was received.
Up until the mid-1800s, artists like Rembrandt were often commissioned by wealthy patrons to paint portraits. Impressionist painters like Manet and Courbet began to paint workers and more mundane scenes, instead of the rich or bourgeois classes. They also used a deliberately simplistic style, with visible brushstrokes and flattened figures. Critics at the time derided these painters as inferior to Old Masters like Rembrandt.
Popular belief is that photography pushed artists into more abstract and conceptual approaches. There is some truth in this — new means of representing the world realistically suddenly exceeded the ability of painting, thus freeing artists up to innovate in different ways.
But Roberts says there’s more to it than that, making an argument similar to Pasquinelli’s about automation: that it was new forms of industrial social relations that prompted artistic shifts in style, not merely technological innovation. Artists’ deliberately simplistic styles were a rebellion against patrons and critics, he argues, as much as their choice of subjects was. An increasingly conservative ruling class wanted to uphold the Renaissance tradition and maintain the past, or even eliminate art. (Sound familiar?) Artistic skill, however, transformed from a technical means of representing reality into a vehicle for departing from realism.
And artists departed further and further: from Impressionism to the Cubism of Picasso and Braque, with its distorted perspectives, primitive forms and new media (like collaging bits of paper onto the canvas) that challenged painting itself; the pure abstraction of Kandinsky and Mondrian; and the ultimate endpoint, Roberts argues, in Duchamp’s urinal. The “ready-made” required no technical skills; the focus of artistic effort shifted to conceptual and critical thinking — the art is in the artist’s choices and deliberations.
Roberts also points to Moholy-Nagy’s 1922 “Telephone Pictures” as the first artworks to be produced in a factory — the artist literally phoned it in, never touching the work. The art was in the instructions — not unlike today’s prompt-driven generative art.
Deskilling then, for Roberts, is not just about artists being replaced by machines (like the camera). There was a commensurate re-skilling in other techniques, and a turning against the establishment. One effect is that this helped to reduce the gender gap, in which women were traditionally restricted from the art academy.
Mechanized thinking
Let’s now apply this to contemporary automation. AI promoters like to tell us that when AI can do our jobs better than we can, this will free us up for more fulfilling, creative work. There’s evidence for this, from repetitive production work in design, to “four D” (dull, dangerous, dirty, difficult) work taken over by robots.
But that’s just on the surface. Let’s take Pasquinelli’s and Roberts’ approach and look at the deeper social relations underlying automation. For manual work, automation has been underway for a long time, with pre-AI robotics already taking up tasks in factories and hazardous environments for example. So here, automation is just a continuation of the industrial era.
The automation of white-collar “knowledge work” is a different story. No one seemed to see that coming, until ChatGPT showed that reading, writing and research were easier to automate than we imagined. Pasquinelli’s argument is that AI doesn’t really imitate human intelligence, but human social relations and labor more generally. Viewed this way, automation is inevitable in all fields, as management will always seek to cut costs and maximize profits however it can.
What’s surprising, then, is that AI can also automate management itself. Here’s where some of the more dystopian scenarios come in — we all start to work for the machines, and AI takes over executive functions all the way up to the governmental level. Some people, like gig workers, already work for AI, essentially.
AI didn’t suddenly become capable of automating knowledge work, according to Pasquinelli; rather, much of what we consider to be irreplaceable “thinking” is actually already systematized, pattern-based labor that can be extracted and mechanized.
A potential solution lies in the fact that such labor — all labor, in fact — is inherently social, existing within relations between peers, employees and managers; and at a higher level between companies, governments and masses of people. Engeström’s analysis goes beyond Pasquinelli’s in looking not only at tools, rules and labor, but adding dimensions of Subject (you), Object (what you’re doing), and importantly, Community — practitioners you work with and who do the same kind of work as you do.
Critique and collab
Zooming down to the individual level, one possible response to automation, drawing from Roberts, is to embrace critical and conceptual thinking. If AI can make hyper-realistic images and video, the artist might shift to deliberately lower-tech, or more material works, or more critical ones. We see many artists doing this. The big question is whether the market will have an appetite for such human-produced work, or whether artists will still need day jobs using generative AI to pay the bills.
Like Manet and Courbet, can contemporary artists afford to distance themselves from the patrons and collectors who pay their way, and the critics those patrons and collectors listen to? More broadly, can workers afford to resist employers, the market and the media? The flaw in my argument is that Manet and Courbet were dismissed at the time, and their innovations only recognized later.
Kayyali pointed out to me that with AI, everyone becomes a kind of editor, curator or conductor, instead of a lone creator. Academic Leif Weatherby argues that automation could lead of a loss of creative skills. What skills, then, might reskilling include?
Weatherby pushes for more liberal arts education. Pasquinelli argues for a “collective counter-intelligence”. Maybe we need to look beyond the “lone genius”. Just as Roberts showed how art became collective labor, workers in all fields might think less about working in isolation and learning new skills for themselves, and more about embracing cycles of deskilling as reskilling, collectively building critical capacity against the establishment. This critical capacity, Roberts argues (drawing from Adorno and Nietzsche) is what actually generates value in the art world. Could it work in other fields? Is this what is meant by “disruptive innovation”?
On the other hand, could management and workers join forces to make automation work for everyone? Economics professor Milena Nikolova found that “automation works best when workers are involved in its design and deployment,” and employers need to help workers re-skill and step into new roles.
Whether or not workers embrace AI, whether or not they join with others in making it work, nothing will stop it from spreading, short of a global catastrophe on the scale of an asteroid strike (I told you I was pessimistic). If we believe the most optimistic scenario — that AI will eliminate all jobs, that means a radically different politics and economy; Google DeepMind’s Demis Hassabis describes this post-capitalist scenario.
In the meantime, as AI becomes embedded in national policy and decision making, and continued automation is baked into our social systems, maybe instead of fighting this tide, we should look again at what cannot be automated, what is uniquely human. Can we see value (monetary or otherwise) in doing what machines cannot, instead of serving them?
