Waning Poetry, Waxing LLMs (2025)

I think LLMs pose a real threat to poetry. That should be obvious on some level. If an LLM—say, an advancing GPT model—can produce a sonnet whose form and attributes are equivalent to John Donne’s, then our appreciation of Donne’s craftsmanship diminishes. Not because it wasn’t impressive for him to craft such poems, but because our awareness of the speed at which a machine can produce them quickly devalues the human effort.

If you’re resisting this line of thinking, consider chess. I bet most of you don’t find Magnus Carlsen’s achievements remarkable, partly because computers now play chess better than we do. I know that argument is frustrating on a couple of levels, including the common defense that we still find weightlifting impressive even though machines can lift far more than we can—but weightlifting’s benefits attract a mate, whereas poetry and chess do not.

Macaulay wrote in 1825 that “as civilization advances, poetry almost necessarily declines.” What he meant by this rather enigmatic statement is something different from what I’ve been discussing about underappreciating the craftsmanship of poetic language. He meant that our interests are changing and that the sensibilities poetry heightens in us—such as marveling at nature or meditating on love—do not survive technological advances. It’s not that nature or love isn’t essential, but that our way of relating to them no longer requires the affectation of poetry. We’re more interested in a nonfiction book about glaciers and their preservation than a poem about their beauty.

There are two reasons poetry might decline. One is that the things poetry once did don’t need doing anymore. The other is that even if something in us still craves poetry, we don’t need it in a civilization-sustaining way. There are too many different things to do—playing pickleball or being a cardiac surgeon—and, frankly, we don’t have time for poetry as time progresses.

LLMs exacerbate both issues. They threaten the first by cheapening the aesthetic pursuit tout court. Art is a tough gig because it doesn’t translate neatly into a localized effort. While attorneys can find work in everything from global international law to small-town torts, artists struggle to survive at the small-town level. LLMs are effectively destroying the small-market wordsmith as a career.1

LLMs also accelerate the second issue: as language becomes more automated, civilization marches further away from poetry. Our techno-futurist impulses get excited about AI in a way the arts no longer do. Horkheimer criticized the pursuit of just facts, arguing that it masked deeper ideological processes. I think, in like manner, that our excitement over advances in technology isn’t really about technology itself—it’s about the new forms of entertainment and experience those advances will provide.

However, before we “let ’er rip” with LLMs and see what kind of hijinks they get up to in creative fields, we must understand that these models are trained to be as defensive of a particular ideology as we want them to be. For instance, they don’t naturally defend a laissez-faire approach to energy policies; those investigations still require a human agent to go into the world and gather the relevant facts.

I’m still uncertain about the downstream effects of LLMs devaluing poetry. I think English teachers and academics in America have been complacent for about fifty years, and the AI revolution is shaking that autopilot into a more forceful response. As I’ve mentioned elsewhere, there are real issues with techno-futurists’ worldview, particularly in how they conceive of knowledge and its purposes.

Poetry is the ultimate meditation. It simply is not for anything in the material sense. Emerson had this in mind when he wrote that the poet “stands among partial men for the complete man, and apprises us not of his wealth but of the common-wealth.” Much of his essay is gobbledygook (such is the nature of the transcendentalists, I suppose), but the assertion is clear: something in our literature reflects our values.2

We should think hard about how those values are passed down. There are real material reasons to be bullish on AI, but we shouldn’t let those concerns cloud our vision of the cultural landscape we’re building. Maybe the local level is the more impactful place to focus our attention. Things are getting pretty unwieldy.