Actions of Judgment against Self-Erasure

23 Jun, 2024

I’m holding this week onto patient amazement at what poet and critic Allen Grossman’s thrown down so far in True-Love: Essays on Poetry and Valuing.¹ Not taking any time for pleasantries, Grossman jumps right into the tension between Theodor Adorno’s famous assertion that it would be barbaric to write poetry after Auschwitz, and poet and Holocaust survivor Paul Celan’s apparent belief that “only poetry is possible” in its wake."² In trying to make sense of the opposing positions, Grossman posits that whatever side you stand on, what Auschwitz indicates is

that something has happened in our time that can be said (must be said) to have changed the meaning and the bearing of cultural practices [and…] that this something that has happened can be understood as new knowledge, an enlightenment, and by enlightenment I mean a recognition added by history to general conscious life that was not there before. This something that has happened bears upon human value, and therefore poetry, because poetry is a way of promulgating human value (the value of persons). Poems are actions of judgment in the light or darkness of knowledge.³

Viewed in Grossman’s manner, I don’t think it would be wrong to say that artificial intelligence, or at least the ways in which access to it is increasingly available to, say, the average office worker who’d rather let ChatGPT write a stupid report than waste his own time doing so, constitutes some level of enlightenment: a recognition that was not there before—and this even if, as in the conflicting views of Adorno and Celan regarding poetry, we don’t really know at this point what it all means. But what I do feel comfortable asserting about AI has everything to do with the last few lines in that quotation, namely, the conviction that it is not “a way of promulgating human value (the value of persons).”

Yes, yes, AI will make calculations for us and fly drones into bad weather and explosive fires and give us readings of quantitative data—and you can’t really argue that that’s a terrible thing. One of the many factors to consider in all this activity, though, is the sheer amount of energy both training and using AI requires; in a recent article devoted to weighing the potential ecosystem benefits of digital tools with their energy consumption, Michelle Nijhuis made the problem clear. In addition to the rare metals and the electricity and the toxic e-waste involved in the very existence and functioning of the devices on which it’s used, not to mention the working conditions of the miners and factory techs that bring those devices into the world, “Artificial intelligence, whose use is growing exponentially, is especially energy-intensive, since each algorithm within an AI system must be trained on enormous amounts of data; a recent study found that training just one algorithm created five times as many emissions as an average US car creates over its entire lifetime.”⁴ Hardly supportive of the value of the planet, much less the persons who live and depend on it—especially when those persons are having decisions of shelter and health determined for them by an algorithm, with predictably horrifying results.⁵

So there is, to put it simply, the biological problem of fostering a phenomenon whose existence depends on the depletion of the very resources that support our own physical being. But there’s also, as you should be unsurprised to find me arguing, the more spiritual—the more poetic—side of AI’s impeding the promulgation, even appreciation, of human value. And really, that comes down to the fact that (as I’ve already argued on at least one occasion), artificial intelligence is meant to do our thinking, even our creating, for us. It might be argued that it’s really there just to get the meaningless tasks off our hands so that we can get on with the good stuff: making art, writing novels, acting. Anyone who’s paid attention, though, to the recent strikes and efforts by SAG-AFTRA and the Writers Guild of America knows it’s also that good stuff that’s under threat by what essentially amounts to a sophisticated robot.⁶ And as should be the case with determinations about medical care mentioned above, that good stuff by nature involves nuanced decisions that require, if they’re to be made well and fairly, time, concentrated effort, and understanding of the insane complexities of a single unique life, much less a community or world of living beings.

Matthew B. Crawford put it well in his discussion of the essential place of language, and its use, in what it means to be human. He tells us about speaking with a man who’d been trying to put together a little speech for his daughter’s wedding. Although the guy at first tried throwing ideas into ChatGPT, he abandoned that route, knowing it “would have been to absent himself from this significant moment.... to not show up for [the] wedding, in some sense.” What the man was doing in his struggle “to encompass in words the elusive truth of his daughter” was providing proof of the more general fact that “things have significance for us, and we search for words that will do justice to this significance.⁷ AI (in particular, large language models, or LLMs) has no truck with significance, merely using statistical probability to offer up a bunch of words often found together in particular scenarios. In furthering the ways in which we’ve gradually handed over our inherently human ability and desire to “articulat[e] life in the first person,” Crawford alleges that AI presents a unique means of working toward our own “self-erasure” and ceding our “subjectivity” to the decisions of a disembodied something with no experience of or interest in what it means to be a reasoning and feeling biological being.⁸

And so Crawford’s article brings us back to Grossman’s assertion about what poems are: “actions of judgment in the light or darkness of knowledge.” In any given situation, we may have enough knowledge or too little—maybe even too much—to help us figure out what to do, or how to explain that doing. As Adorno seemed to think, our new, frighteningly exhaustive post-Holocaust knowledge about the things some humans are willing to do to others means that expressing ourselves via formats thought to attend above all else to soothing aesthetic qualities was so offensive it shouldn’t even be tried. Celan, on the other hand, thought that in the wake of what had happened in the camps, in that situation of at-best incomplete knowledge, language was the only “thing [that] remained reachable”—but even so, it still “had to go through the thousand darknesses of murderous speech…. It gave me no words for what was happening, but went through it.” In this new world, Celan struggled “to write poems: in order to speak, to orient myself, to find out where I was, where I was going, to chart my reality.”⁹ Language had been eviscerated and brutalized, Celan seems to be saying—but even so, its use, and in particular, the unique absences and incompletions and strange formulations that are only possible in poetry, was still the only way forward in attempting to make sense of anything at all, including especially one’s own place amid the wreckage. Straining “to articulate,” as Crawford said, “life in the first person” was not the decorative distraction Adorno seemed to believe was characteristic of poetry, but was instead the only way Celan saw of holding onto his human place in a world that had made it very difficult to determine what human being or value even meant anymore. It’s precisely what Crawford has in mind in his examination of humans caring about—needing—significance, and LLMs being totally indifferent to it, even incapable of grasping what it could be.

It’s all returning me to a recent bittersweet experience: a friend who does make unconcerned use of ChatGPT, knowing my aversion to it, ordered the thing to write a haiku for my birthday. Now, if you know anything about haiku beyond its most simplified version—a poem of three lines, where the first and third lines consist of five syllables, and the second line of seven—you’ll know it’s a doggedly difficult form to master, if such a thing is even possible. Throw in the requirements for metaphor- and simile- and adverb-free images, particular references to seasons, the cut at the end, the sparing punctuation, etc., etc., and you quickly begin to realize you could spend your entire life trying to write a merely acceptable haiku without coming anywhere close to succeeding.¹⁰ But it’s only by submitting yourself to that process of aiming for the impossible—a long, frustrating, beautiful experience of slow discovery and self-articulation through repeated failure that shortcuts like ChatGPT are meant to eliminate—that you begin to appreciate what an amazing, amazing and uniquely human, thing this poetic form can be.

So all of that gets loaded into the one particular poetic form my friend asked the bot to create for me. To my great sense, though, of triumphant glee, upon reception of a human demand to produce something in said form, ChatGPT crashed, unable to do the job. That outcome was likely a fluke, and I’m sure the tool has churned out plenty of average 5–7–5 combos. But ask an algorithm to produce something up to snuff for a human being wedded to the value of significance—ask it to undertake an action of judgment, and not just sort through information—and you’ll be left with digital, barbaric wreckage. Let’s hope there’s an articulate human within reach to pick it all up and repair it with all the inefficiency, messiness, and wrangling that action deserves.

Subscribe to Off-Modern Onions!

<sup>1. Allen Grossman, True-Love: Essays on Poetry and Valuing (Chicago: University of Chicago Press, 2009).↩
2. Emphasis in original. On p. 175, n1, Grossman provides us with what Adorno actually said: “To write poetry after Auschwitz is barbaric and this damages the very knowledge of why it is impossible to write poetry today.” And I only describe Celan’s own opposing view as apparent because he never stated said view explicitly; the phrase comes from critic Christopher Fynsk, and is quoted in Grossman, 175 n2.↩
3. Grossman, 2. Emphasis in original.↩
4. Michelle Nijhuis, “The Digital Planet,” The New York Review of Books, April 18, 2024, https://www.nybooks.com/articles/2024/04/18/the-digital-planet-environmentalism-from-below/.↩
5. Jesse Bedayn and The Associated Press, “AI is already screening job resumes and rental apartment applications and even determining medical care with almost no oversight,” Fortune, March 5, 2024, https://fortune.com/2024/03/05/ai-screening-job-resumes-rental-apartment-applications-determining-medical-care-almost-no-oversight/.↩
6. For example, see Dawn Chmielewski and Lisa Richwine, “Hollywood actors secure safeguards around AI use on screen,” Reuters, November 9, 2023, https://www.reuters.com/business/media-telecom/hollywood-actors-secure-safeguards-around-ai-use-screen-2023-11-09/. There’s a connection here, I think, to Marx’s hopeful and never-achieved assertion that machines could/would save us all a lot of labor, so that we could end up with a lot more time to cultivate and enjoy ourselves—but instead, we just find ourselves working/being worked all the harder to keep those productivity numbers up.↩
7. All quotations in this paragraph are from Matthew B. Crawford, “AI as Self-Erasure: Humanity’s will to disappear is being installed in the omni-operating system,” The Hedgehog Review, June 11, 2024, https://hedgehogreview.com/web-features/thr/posts/ai-as-self-erasure. Emphases are original.↩
8. Crawford notes that the problem didn’t begin, say, with algorithms ranking news stories; he cites among others the observations of Alexis de Toqueville and Søren Kierkegaard on mass society, and newer discussions on collective decision making in the office, education being viewed as the mere passing on of job skills, and the functioning of dating apps as different indicators of the ease with which we give up our ability and willingness to make judgments.↩
9. Paul Celan quoted in Grossman, 9.↩
10. For a good, but not complete, discussion of the intricacies of haiku creation, check out Julie Bloss Kelsey, “New to Haiku: How Is Writing Haiku Different from Writing Prose Poetry?” The Haiku Foundation, October 10, 2021, https://thehaikufoundation.org/new-to-haiku-how-is-writing-haiku-different-from-writing-prose-poetry/.↩</sup>