what I dislike about AI isn't the tech (and why I like Ellulian 'technique')

Last Thursday, I listened to a recent episode of The Vergecast during my morning bike commute. The episode featured Paul Ford talking about his recent experience with Claude Code, and I was genuinely surprised to find some of his comments resonating with me. It helped that Ford wasn’t uncritical about AI (though certainly not as critical as I would have been), but some of it was just that I recognized some of the thrill that he was describing of using tools and resources to learn how to solve a problem. In fact, I found that thrill so contagious that a passing comment he made got me to spend some time once I got to the office converting my Twitter archive into a CSV that I could finally import it into the Day One journaling app that I use.

Of course, I did all of that work by hand, because I’m still at a point (and hope to stay at a point) where I refuse to use generative AI tools to contribute to my work. Nonetheless, listening to the interview really got me thinking about what it is about generative AI that bothers me so much—and it’s pretty clear that it’s not the technology itself. For all of their hallucinations and similar issues, image generation tools, LLMs, and the like are technically impressive! Just in terms of what they can do, they are pretty neat technological developments, and for all of my grumpiness, I’m happy to concede that.

So, what makes me angry about generative AI isn’t the technology itself, it’s all the baggage that it brings with it. It’s the extractive nature of how the models are trained. It’s the scattershot, unreflective way that the technologies are being rolled out. It’s the insistence that this is the future and that everyone needs to get on board. I wouldn’t necessarily be opposed to a carefully and consensually trained model that was applied in specific circumstances without being imposed on anyone—that could be a really neat application of these underlying technologies. That’s not the generative AI that we’re being presented with, though, and my abstract appreciation for “neat tech does neat things” doesn’t survive contact with the status quo.

Thinking about all of this reminds me why I appreciate the work of Jacques Ellul as a theoretical framework for thinking about technology. (Today, coincidentally, marks the first anniversary of my having finished his The Technological Society). While Ellul is often (and not incorrectly) described as a critic of technology, it is more accurate to describe him as a critic of technique, a term with different connotations. Some of these differences are due to linguistic and cultural differences between the Anglo-Saxon world and Continental Europe, but Ellul also had an idiosyncratic way of using the term that distinguishes him from other French (and Continental) writers.

In short, technique is more than any given tool or technology. First, Ellul insists that the entire technological system—the (translated) title of another of his books—be considered together instead of as individual pieces. Second, Ellul emphasizes that societal trends and attitudes—chiefly, an unyielding insistence on efficiency and efficacy—are part of that technological system. In fact, it’s easy to miss among his fierce criticisms that Ellul is not necessarily opposed to individual tools and technologies. His critique is most specifically oriented at that societal trend, at the insistence that one must adopt all new technologies because they are more efficient. Here’s a taste of that in the posthumously published Théologie et technique, where Ellul argues for an ethic of non-power, in which people deliberately choose not to do everything that they’re capable of and are more deliberative about what makes sense (again, not just what is possible—it’s a very Ian-Malcolm-in-Jurassic-Park ethic):

Le système technicien ne peut strictement pas supporter une attitude de vie de non-puissance, ce serait sa ruine : il suffit de penser que l’on ne choisirait plus, pour consommer, ce qui est le plus rapide, le plus efficace, le plus perfectionné. Je ne dis pas que l’on choisirait systématiquement le moins rapide, le moins efficace, etc., mais simplement on cesse d’être intéressé par ces qualités.

Here it is in my rough translation

The technological system strictly cannot tolerate an attitude of non-power toward life, for it would be its ruin: It suffices to think that one would no longer choose, when consuming, what is the fastest, the most efficient, the most perfect. I do not say that one would systematically choose the slowest, the least efficient, etc., but simply that one would cease to be interested by these qualities.

Ellul’s technique is helpful for me in understanding my frustration with generative AI. It’s not the tool itself, it’s the broader system that surrounds it. If generative AI could be separated from the societal attitudes of “what can be done must be done” and “what seems more efficient is necessarily better,” I think I could get on board with the tool in a way that I am currently not. However, 30 years after Ellul’s death, I feel like generative AI more than anything else shows the utility of this theoretical framing of this, and other, technologies.

• generative AI
• The Verge
• Jacques Ellul
• bike commuting
• Claude
• theory
• Twitter
• journaling
• Day One
• digital labor
• Théologie et technique