The AInguish of Influence

Creators have always had a tense relationship with their influences, but generative AI threatens not just the meaning of art, but the very nature of the self.

And so but 1996 was the Year of David Foster Wallace. More specifically, it was that summer we all read Infinite Jest, grad students avoiding real work by tag-teaming and overlapping in a communal reading of a couple of increasingly ragged copies of Wallace’s thousand page metafictional masterpiece. We would read all day and then talk about it over drinks at night, trying to parse the actual plot while trading off favourite lines or references or anecdotes.

The baroquely constructed dystopian-comedy plot aside, what was exhilarating about Infinite Jest was just the style of the writing: the run on sentences and the interior monologues and the footnotes and the footnotes to the footnotes and fourth-wall breaking and just the relentlessly and frequently exhausting pyrotechnic approach to spelling, grammar, structure, and voice. At least two of my friends soon contracted the habit of answering the phone with a drawn out “mmyellow?”, mimicking the style of one of the main characters, while I tried to jump start my then-nonexistent journalism career by trying to write just like DFW in an ill-advised column for the student newspaper.

It all seemed fantastically new and fun and breathtakingly original. Only later, when I got around to reading Wallace’s acknowledged influences (DeLillo, Pynchon, and Gaddis especially) did it become obvious just how much he had borrowed, adapted, or outright stolen. (One character from Infinite Jest, who wanted to be a sports reporter and would narrate matches from the sideline into his fist like he was holding a microphone, is straight up ripped off from an early and relatively obscure DeLillo novel called End Zone).

It wasn’t that Wallace no longer seemed like an incredible writer and intellect – he absolutely was – but rather, just that he seemed more plainly derivative. Not in a bad way, just more obviously human. And when it came to all of that writerly gymnastics, it was clear that he was doing more than just having fun with the reader; he was desperately trying to outrun (or at least out-write) his influences when at all possible, and acknowledge them when not.

All of this is detailed in probably the best critical essay written about David Foster Wallace, A.O. Scott’s “The Panic of Influence,” published in the February 2000 issue of the New York Review of Books. As Scott writes in his omnibus review of DFW’s work up till then, “It’s hard to think of another writer of any generation who has written more prolifically about the obstacles to writing, or who has lampooned the self-dramatizing frustrations of the creative process with such inexhaustible, maniacal conviction.”

Wallace’s central problem, Scott notes, is that he suffered from an extremely acute case of what the literary critic Harold Bloom, in a 1973 book about poetry, called “the anxiety of influence”. As Bloom argued, every poet takes his or her inspiration from another poet’s work, and as a result, runs the risk of writing derivative or “weak” poetry. But because every poet wants their work to be seen as original or visionary, the mere existence of influences generates a deep sense of anxiety in the writer. For the relatively few accomplished “strong” poets, this anxiety results in truly original work that stands for posterity, while the rest are doomed to obscurity.

Wallace was consumed by the desire to avoid obscurity and establish his bona fides as a strong poet. But he was also obsessively aware of how much he drew upon his predecessors; he wore his influences thickly, openly celebrated them, and went to a great deal of trouble to let the reader know that he knew that you knew that he had stolen a riff or a line or a schtick, while still trying to shrug off the influences and say something clear and novel, though clearly footnoting anything that might strike the careful reader as ever so slightly borrowed. More often than was healthy, this resulted in writing that was a complete mess, a funhouse-mirrored ouroboros of self- and self-self- and self-self-self-reference.

You don’t have to be a poet or novelist to suffer from the anxiety of influence. Creators of every sort, from film to painting to music or just anyone who deals in the manipulation of cultural symbols more generally, is always in a complicated and often tense negotiation with the past. Indeed, that negotiation forms the very core of the idea of the self, and in this identity-obsessed age, the struggle between identity (your authenticity or uniqueness) and influence (your heritage) gives rise to a deep paradox.

It goes like this: On the one hand, we prize originality, we celebrate those who are able to go their own way, do their own thing, resist the conformism of the masses. But on the other hand, the very concept of an identity involves a commitment to predictability. Think of your favourite singer or band: if you hear a new song, you can usually tell it is them within a bar or two. Artists have a distinct style, as do filmmakers. But forget the creative arts, and just think about your relationships: someone who understands you the best is the person who already knows what you are thinking, or what you are about to say, or how you will respond to a given situation.

We try to resolve the paradox by convincing ourselves that we are all unique, true originals, authentic selves, and that any predictability simply comes from people becoming attuned to our quirks and character. But what if the predictability goes deeper?

A recent paper in computer science reported on the result of an experiment in using generative AI to simulate the “attitudes and behaviours” of over a thousand real people. The researchers did extensive interviews with the subjects about their lives, and then fed the results into LLMs (the large language models that are the foundation of generative AI) to create “generative agents”, i.e. simulations of the real individuals. The results were pretty impressive: The agents were able to replicate participants' responses on the survey at least 85% as accurately “as participants replicate their own answers two weeks later, and perform comparably in predicting personality traits and outcomes in experimental replications.” Put simply, if you train a LLM with someone’s attitudes and behaviours, you can create an agent that will credibly mimic their identity. And keep in mind, when it comes to AI research, we are in the paleolithic era.

Yet AI has already made deep inroads into the creative industry. For all the worries (or maybe it’s hope) about the coming AI sludge and predictions of the inevitable collapse of AI models that end up just consuming their own garbage, people are already using it to create original, interesting, and in some cases beautiful works of art – from painting to poetry to literature to film. We are not far from being in a position where the question of whether something was “created by AI” is no longer an interesting or useful question to ask.

Does that mean the end of human creativity? Not in the least. If anything, it will open a new space in the culture for more considered and meaningful forms of pure human-powered engagement and creativity. As the artist and writer Austin Kleon boasts at the end of his excellent substack, “This is a hand-rolled, ad-free, AI-free, anti-algorithm newsletter.”

It does raise hard questions, though, for the very idea of the self. If, as is looking likely, we are all going to be effectively replicated by autonomous agents programmed on our own behaviours, what does it mean for identity, agency, free will? Are we reducible to mere patterns, predictable and formulaic? Are we all, deep down, just glorified auto-complete, algorithms trained on the Big Datasets of our DNA and our childhood?

I increasingly think that is more true than not. The good news is that maybe my kids will be able to hang out with digital me, long after I’m gone to dust. Want to know what pops would have thought? Let’s just ask ChatGPDad.

Maybe that sounds dystopic, like a bargain-bin episode of Black Mirror, but honestly I kind of like the idea. And how different is it, ultimately, than the endless hours we already spend with Aristotle or Shakespeare or Dickens or Austen or Rothko or Kahlo or Prince… or David Foster Wallace, who killed himself in 2008 at 46, but whose mind I can visit anytime I like, with a short stroll over to my bookshelf. We interrogate, argue with, and learn from the past all the time; our digital future is in many ways just a more democratized and interactive form of what we already experience.

At some point, given sufficient time and perspective all of these influences just sort of blend into one big cultural inheritance, the great human mystery experiment, and the anxieties over any one contribution seem small and insignificant.

---

From the X-Files:

• ‘Zine scene chronicler Broken Pencil magazine is over and done with (also, Hal has a Substack)

• Vice founder Shane Smith has a new show investigating “disinformation”

• The inevitable backlash against the backlash to “Do They Know It’s Christmas”

• It seems that every microgeneration has to discover Mark Leyner for themselves

• The photo of David Foster Wallace at the top of this newsletter was generated by Dalle 3