The (generative) AI discourse is nothing new

There is something darkly amusing to the panicked reactions around the impact generative AI has on labor. Not because any of it is funny, hell no. Not because it's not a real problem, far from it. But because the fear I see from writers and illustrators feels so... familiar. The critique so early-stage. And that's because I'm a handweaver and a crafts history nerd.

Let's wind back for a second and define our scope somewhat. We're in a full-blown boom of "AI" hype. I put "AI" between quotes because artificial intelligence - that's what the acronym means - is a gigantic field with a dazzling variety of techniques, but the discourse is typically centered on a narrow subset of user-friendly tools leveraging machine learning in specific ways, which I will call "generative AI tools" here. As a somewhat old-school programmer I've always found the approach inelegant due to its brute-force nature, but that's not the topic today. Also as a somewhat old-school programmer, I am not even going to dignify laughable fears of basilisks and Skynet-esque end of the world scenarios with even a second of brainpower. I will also leave aside the true cost of those technologies notably when it comes to energy and water. My focus is solely on how the related labor issues and quality-centric counter-arguments echo centuries-old crafts discourse.

Generative AI tools, applied to crafts such as (digital) illustration or writing, allow virtually anyone to spit out a large amount of seemingly-passable works in very little time. This, in an economy when quickly churning out "content" is valued more than quality or the livelihood of creatives, is obviously a direct and dire threat to the continued employment of many, many people - which has led to tremendous hostility towards anything vaguely associated with "AI". In turn, generative AI enthusiasts (who often stand to make significant financial gains from the ongoing boom) tend to mock those fears as those of luddites. They're not wrong, if only because the luddites very much had a point: note how much of the story is about wages and food prices, not hating machines. But if we take that craft link so handily given to us, and dig deeper into it, what do we find?

One of the things we find is Walter Benjamin's The Work of Art in the Age of Mechanical Reproduction, written in 1935. You can read the whole thing here or stick to the Wikipedia summary. The essay isn't very long but it is dense, complex, and touches on many ideas. Among other things, it highlights that reproducing and therefore "cheapening" works of art is as old as art itself, and discusses various social and political functions of artworks both original and reproduced. Like any good work drawing from Marx, it of course speaks of revolution. Technology as a threat of sorts to the nature of art itself is, indeed, not new at all. Neither is thinking about it by bringing together philosophy and politics.

One common seemingly-philosophical argument from creatives faced with the threat of technology is that the handmade has inherently superior value. Newcomers to the topic tend to peddle platitudes, and that is not entirely their fault: few stop to think about what "handmade" even means, and why it is supposedly valuable. One could argue that many of the digital illustrators suddenly beamoaning excessive computer assistance were once themselves painted as cheaters by traditional illustrators. I am not, however, here to make fun of anyone: I am here to give you the sword that will cut that particular Gordian knot. That sword is a fantastic essay by David Pye in The nature and art of workmanship, titled Is anything done by hand?. You can read the entire essay online, it's fairly short and considerably more readable than Walter Benjamin.

Its key insight is halfway in: "Is it not time to give up and admit that we are trying to define in the language of technology a term which is not technical? 'Handicraft' and 'Hand-made' are historical or social terms, not technical ones." Because that is what this is all about, isn't it? It's not about how much computer by volume was involved in the making of a work. It's about something more diffuse and yet so very tangible: human involvement and expertise, human control beyond curation alone, continuity with a tradition, creative works rather than content. Something with intent and meaning, not just pixel slurry. Something that doesn't suck. That is what the takes are grasping at when defending the handmade. It's not about the tools. Ironically enough, this means that a sufficiently bespoke, sufficiently curated, sufficiently interesting generative AI tool trained on an ethically sourced dataset would be just fine. This is, in fact, happening in audiovisual production pipelines everywhere, you just can't tell from the result. It's a good thing! Automating grunt work is a good thing. (This could be its own essay. Mechanization was supposed to free us from labor.)

We tend to draw a hard line between the artisan and the factory, between the hand-tool and the machine-tool, between the handmade and the industrial. Reality is more of a messy spectrum. The Shape of Craft by Ezra Shales is an entire book on this very topic - and where I found David Pye's essay to begin with. Ezra Shales also has a number of presentations online if you'd rather watch a video. The book is a deep exploration of craft via a number of case studies, for example about the complex hand-work of pouring ceramics in a factory. My own version of analysing those apparent contradictions will, of course, focus on weaving. At what point does a loom stop being a hand-tool, and become a machine-tool? Is it when you add shafts that avoid hand-picking threads? Is it when you add a fly shuttle that lets you pull a handle rather than throw the weft by hand? Is it when further automation lets you simply push pedals like you are riding a bike? Is it when you add a motor? There is no right answer. Because it isn't really about the tool. It is about history, about social dynamics, and about who gets paid.

To sum this all up in a single tweet-sized statement? You're not mad at computers, you're not even mad at generative AI. You're mad at capitalism (and tech bros). And textile history can show you the path forward: labor organizing.

Part of organizing is admitting that the world changes, though. That is why among the many takes about generative AI tools, one that stuck with me speaks of Zero Trust Homework. How does homework keep existing when generative AI tools become widespread? How does homework remain useful and become better (perhaps for the first time in a while) in this new information age? Maybe it must switch gears and focus on honing different skills. Maybe the future is one with more curating, more checking, more refining than grunt work. Maybe that can be made not to suck.

Sourcing notes:

• I can't recall how I ended up at Gustavsbergs Konsthall for a talk by Ezra Shales back in April 2019, but I believe that's where I first heard of The Shape of Craft, picked it up, and had my mind blown.
• As mentioned, David Pye's essay is brought up in The Shape of Craft.
• I heard of The Work of Art in the Age of Mechanical Reproduction via someone snarking on Twitter at some point. It was probably about NFTs. Remember those?
• I have been doing textile crafts for over twenty years and weaving by hand semi-seriously since 2015. I am especially interested in the history of loom mechanics, which resulted in a bunch of study visits in Lyon (2016 and 2017) and Paris (2018), plus a two-week class in Florence punching actual cards to weave on antique looms (2017).
• My point with this section? My ideas here are mostly shaped by... literal years of thinking about the topic. Welcome to the fight. It has been raging for decades, and we did score some victories even as the world changed. You're not alone.