Ellul, nuclear weapons, and generative AI

One of the most interesting recurring themes in Jacques Ellul’s writing is one that contrasts reality (or facts) with truth. As Ellul distinguishes them, facts are what are and—implicitly—what must be conformed to, whereas truth is what ought to be. Ellul’s The Humiliation of the Word explores this distinction at length, but it crops up in plenty of his other writing. In fact, I’m currently reading his Présence au monde moderne (or rereading it, depending on what one considers reading the original French after reading the English translation last year), and I’m delighted to see that he makes this distinction as early as this 1948 book.

I’ve previously written (though in French) that I find this distinction helpful for interrogating the idea that “well, since AI is here, we have to learn how to use it correctly.” I tend to disagree with this take, but I have a certain amount of appreciation for it. People I deeply respect have made this argument, and as much as I strive to be more radical, my default posture is one of pragmatism. Furthermore, in my professional life, I’ve taken steps to approve and encourage classes in my department that are focused on AI, basically operating off of the “if it’s here, we should teach about it” principle.

Yet, I’m still not satisfied with the idea, and I think this passage from Ellul about nuclear weapons nicely illustrates why:

Those who question the value of the fact incur the harshest of all reproaches today: they are reactionaries, they want to return to the good old days, and so on. . . . Those who level this reproach do not realize that such doubt is perhaps the sole revolutionary attitude that is currently possible. Even so, we need to know the reason why we refuse to bow before the fact, for the surrealistic approach does not seem desirable.

The atomic bomb provides quite a striking example of this religious authority that the fact carries. In the face of this discovery, this instrument of death, humankind retained the possibility of not using it, of not accepting this fact. But this question was not even posed. We found ourselves before a fact; thus we had to accept it. And from that point on, the questions asked were “secondary.” Who will use this weapon? How will its control be arranged? What will be most expedient: to use nuclear power for war, or for peace? How can the economy be organized around nuclear power? and so forth, and so on. At no time was the problem posed of knowing if it was good or evil to embark on this path, and this was because the fact today seems to be beyond good and evil.

Now, Ellul is prone to a bit of rhetorical exaggeration, and I’m willing to concede that there are differences between LLMs and atomic bombs. Even with those caveats, though, I think this passage remains effective for challenging a simplistic “so long as it’s here, we might as well do something with it” defense of AI. As he argues, the “fact” of a particular technology is not an ethical justification of that technology. I’d like to hear more reflection from those who make the “well, it’s here” defense of AI on what the ethical harms of the technology are and why it’s worth using/promoting/teaching anyway.

• Jacques Ellul
• generative Ai
• technological determinism
• reality vs. truth
• The Humiliation of the Word
• Presence in the Modern World
• Présence au monde moderne