For last month's launch of Deluge's new A.I. novelette AMOR CRINGE, we hosted a Zoom conversation between author K Allado-McDowell and literary theorist Peli Grietzer. Below, see a transcript of their talk, which has been edited for length and clarity.  

 

ON GPT-3 AND FICTION WRITING

K Allado-McDowell: I was in Montauk over the summer, and I had this office upstairs at the house where I was staying. I kept telling my partner, “I’m writing this evil book about cringe up there.” It quickly turned into fiction, partly because that was where GPT-3 was going. I'm not quite sure why, but this is the way I like to work with GPT-3: I like to allow it to pitch a text or a set of ideas in a direction that I'm not expecting. I find that helpful.

[Amor Cringe] tells the story of an influencer who lives in the basement of a TikTok house and is seeking transcendence and the divine and having a hard time finding it through the structures that are afforded by our culture. It's a lot about the subjective effects of social media and a psychological profile that emerges from being self-aware, but not quite self-aware enough to transcend your own self-awareness, and being paralyzed by that, and the kinds of access that gives you to transcendence. That's a pretty highfalutin way of putting it. It's actually quite absurd, bawdy, funny, grotesque, and at times, horrific. It was a way for me to explore a lot of things I haven't been able to explore in writing. My previous book, Pharmako-AI, very utopian. It’s theoretical and personal, too, but this book was like, Let me just stick my nose in all the cringiest things in culture right now.

If you’ve read Pharmako-AI, you know that I wrote it in a way where the voices were clear and distinct. They are even typeset differently. The whole thing is like a recording of a conversation, in that I didn't change what came out of the model after a certain point. If I set something in place and responded to it, then I would leave it in place and not go back and edit. With Amor Cringe, I decided to go completely freeform. I let myself do whatever I wanted—copy, paste, rewrite sections, change it all around—and be creative with it, as opposed to trying to be extremely systematic. There are sections that I know for sure I wrote, sections I know for sure were generated, and there are sections where I can't remember. That process was much more fluid. I think because it’s fiction, it needed to be that way. GPT-3, in my experience, is pretty bad at writing fiction. It’s got a short contextual window. It’s 2,000 tokens, which is about 2,000 words, and it doesn't really remember. If you put the gun in the first act, it shoots it in the first act. It doesn't wait for the third act, so you need to play with it and push it a little bit.

 

FIGURES IN A LANDSCAPE YOU CAN INTERACT WITH

K Allado-McDowell: In Benny Shanon's The Antipodes of the Mind, he describes a survey of hallucinatory effects: often, they begin with patterns that turn into shapes that turn into figures that turn into figures in landscapes that turn into figures in landscapes that you can interact with. This aspect of hallucination is present in Amor Cringe.

I knew that I wanted the book to explore feelings of shame related to narcissism. I had a heuristic which was to always opt for the cringe option. If there were options on the table, I’d pick the one that was the most cringe. I would subject the character to the most cringe-inducing experiences that were available in the narrative.

Certain things were seeds. In the beginning, “In the TikTok house the boys are dancing” (Amor Cringe, p. 9) — I wanted that to be a scene. “I started dating a guy I met at Erewhon” (ibid, p. 53) was also a seed I planted, I wanted to see what would happen with that. There were certain things that I knew explicitly I wanted to be scenes, but then there were things that came out. There’s one section that illustrates this well:

A few months before the fire, she’d escaped the family for a medical appointment. We met outside a Duane Reade in SoHo. She picked up her prescription. As I waited, I cataloged the women on the product labels. Gold locks glimmered on bottles of purple shampoo laced with sulfates and parabens, destined for the ocean. She walked up holding a white paper bag and looked at the rows of plastic bottles. “How can you not be an atheist?” she said.

It amazes me, but it’s actually true. I went to church until I was nine, and I still go sometimes and pretend to pray for this dead saint or that dead martyr or the dead priest who used to live downstairs. I sit in church, completely believing that I’m one of them. I’m deceived by all of it, the fake historicity of the architecture. I listen to masses in Latin as if they were historical reenactments. It’s been the same every week for millennia. It’s a contract, a collective obsession. (ibid, p. 19)

 

I wrote the scene in the Duane Reade with the shampoo bottles. At this point, I can’t remember if the line, “‘How can you not be an atheist?’” came from GPT-3 or from me. I feel like it came from GPT-3, but I honestly have lost track of that. But the part after, when the narrator says, “It amazes me, but it’s actually true. I went to church until I was nine…” and describes this self-awareness about being deceived by a ritual in a church, that part came from GPT-3. The story of seeking spiritual transcendence and trying to find a way of having an authentic religious practice in a world where you feel like you should probably be an atheist, that theme, in many ways, emerged from GPT-3’s responses. One thing I found writing with GPT-3 is that often, through its pattern-matching ability and its reflection of the patterns in the text, in comparison to my own incomplete understanding of my motivations as an author, it produces the effect of seeing your subconscious on the page. 

The organic process surfaces something that you’re subconsciously putting into the text and gives you a reflection of it, and then you can choose to carry that forward or not. Religion becomes a big theme in the book—the character trying to go to these different churches and being dissatisfied by the clothes the people wear, the architecture, or the lack of charisma, and then ultimately finding a church, only to play out their pattern of narcissistic sexual predation and consumption in that space.

It was definitely collaborative on a level because these themes kept coming up through the process. I didn’t have a plot in mind. It was more like playing a role-playing game. I don't know who is the game runner, really, if it’s me or the GPT-3, but somehow these things would happen. There's one scene where there are people coming to update the building that the characters live in because the water is rising. It’s in Malibu and the waves are always beating against the basement of the house. They need to do some construction on the building. These people invite the narrator to play a gig, a DJ gig, which is something that came out of GPT-3—visiting and inviting them to play a DJ gig. They get super fucked up at the club and they totally blow their set. They don’t even mix out of the first song because they’re too fucked up to know what’s going on, and then it just spirals from there. In a traditional narrative, this would be the crisis that produces a change in character and then gets played out, so it’s hard to say how these structural elements emerge. In some sense, I was reproducing elements of traditional novels that I knew would work.

 

LIKE A SYNTHESIZER

K Allado-McDowell: Music appears in both Pharmako-AI and Amor Cringe as an important subject. The narrator in Amor Cringe is a musician. If you don't mind, I would like to read how they describe their music:

I always said I made music because I couldn’t do anything else. Nothing else seemed like it mattered. Sometimes, though, I would be clicking a mechanical pencil or spinning a fidget spinner shaped like a UFO and realize that I got almost the same satisfaction from the repeated spinning of the ball bearings between my fingers or the compulsive releasing of the pencil lead, and it made me wonder if my music was garbage. I mean, it probably was. I knew nothing about music that I hadn’t learned from drugged, visceral immersion in parties and what amounted to random paths through Ableton’s dropdown menus and sliders and frictionless knobs. You could study the theory of music or audio synthesis, but what difference did theory make when what you wanted was an elusive combination of sounds, like an ever-changing key to the brain stem? Why was one song sexy and another driving? When I mashed the buttons on my controller, I sometimes felt like a monkey in a lab or a child with a Simple Simon. My aunt used to play the oboe, and I’m pretty sure it never made her feel that way. (ibid, p. 45)

 

There’s a cringe aspect of making music with a computer: feeling like you’re just touching buttons, you’re disembodied, anyone could do this, and this is probably trash. Part of this book is about embracing all those things, too, and asking, What can be found in that? In Pharmako-AI, music appears as a structuring principle. Here, it’s an object that is a source of shame and narcissism. 

I have set some poetry from Pharmako-AI to music. I'm working on an opera project that will have some AI-generated music in it. On a deep level, music is the way I think about reality, so it’s hard for me not to bring that into play here, but it’s also the aspect of the tool being an instrument, in that it, however large, has a fixed set of capabilities and what you do with it [is] your voice — is the music that you make. Before I wrote Pharmako-AI and Amor Cringe, I had written two half-complete science fiction novels that I couldn't finish. I think I know why. There were some plot problems, but it was also not thinking of writing as playing an instrument. Maybe the metaphor is flowing in the direction from music to writing, but it’s starting to flow the other way, too.

Peli Grietzer: Something that’s been surprising me in this conversation — and interestingly it’s also something that surprised me a few years ago talking to Holly Herndon about the Spawn project and the album she did with it. In both cases, I expected a lot of the passion for the artist to be around shamanic relationship with AI, where the goal was to try and manifest in the work of art whatever it is that lies in AI as a mathematical or social object, or as a totality of language, materially embodied. In both cases, once we started talking, I’m surprised that both of you talk about AI like how people talk about a synthesizer in terms of, This thing has very interesting capabilities that can seamlessly integrate to the things that I'm already interested in achieving as an artist. Basically, where the air enters into the work of art genuinely as a tool is too diminishing, but as a process, rather than as the implicit object of the writing. Holly doesn't think of the music she did with Spawn as being about Spawn or exploring the essence of Spawn. It also seems like you're not necessarily thinking about what you’re trying to do as communing with the collective unconscious or the collective being of language embodied in this machine, but rather working with it as a cultural or technical source of inspiration of technical means for realizing ideas.

K Allado-McDowell: Maybe that’s just artists being artists. There's a certain degree to which you have to be like, I'm playing this thing and I am making this music or this text for a specific purpose. There’s lots of good work around what is going on in the model, trying to understand it as an object and a social phenomenon. That’s what a lot of critical AI and the AI ethics conversation has revolved around. To me, being very familiar with that conversation, it’s not as interesting to unpack, artistically, for my purposes. When I think about Amor Cringe, a lot of what I wanted to explore was the relationship of cringe and spirituality, the resurgent desire for spirituality, and what it means to sit in the cringe space for a long time. None of those are necessarily implicit in GPT-3. Pharmako-AI does address models or the social function of AI more directly. And because Amor Cringe is fiction, I wanted it to function as a normal piece of fiction.

 

IDENTITY IS CRINGE

Peli Grietzer: Let's talk about cringe. Perhaps a little bit in relation to AI, but not necessarily. You’ve been writing recently on Twitter about cringe as sort of a metaphysical, epistemic social concept, in a way that harks close to the problem of irony as understood in 19th century German philosophy. Irony was understood as this idea that there seemed to be many perspectives through which we could understand the world, in terms of political, metaphysical, moral, and aesthetic matters, so how could you, as a person, speaking on knowing that one’s knowing is finite and your particular empirical cell, rather than the horizon point or end of thought, how could you acknowledge your constraint and contingency and still be able to say something or to speak at all? For the Romantics, a slightly big idea was to use irony in a way that didn’t necessarily cancel what they were saying, but turned their particularity into a mirror of infinity of other particularities that there could be. In this way, one transcends one’s fallenness as an empirical being, rather than an infinite being, weaving in that acknowledgement of the infinity of other possibilities into whatever it is that you're doing at the moment. Of course Hegel was very upset by this because Hegel thought that this was a cop out from ever having to plow on in dialectical struggle toward actually reaching the absolute point of thought. 

Do you think that leaning into cringe is the same as the sophisticated philosophical irony that has been the previous famous solution to the problem of self and other and the problem of infinite perspectives, or is leaning into cringe something that speaks to something more specific about our current predicament with the problem of the infinity of other perspectives and that withering terror that one can experience?

K Allado-McDowell: In the way you’re describing irony, there's a teleology that’s being questioned, but also being resolved. It's this idea of reaching absolute understanding, or failing to and finding a way to deal with that teleology. What I am proposing in the Notes on Cringe that I posted on Twitter, is that if there is a teleology, it’s more like a Buddhist form of enlightenment. There are these mechanics that are involved in order to cringe. You need to observe an other, which requires maintaining a model of them, and then modeling their own embarrassment, or lack of embarrassment, and then responding to it with your own emotional response of embarrassment, or shame, or however we want to put that. The idea there is that you see somebody do something, and in this immediate, emotional, limbic response you have a feeling of anxiety, alienation, or shame on their behalf. You’ve quickly constructed a model and put yourself into it and resonated with that, so that absorption of the model implies that your own model is also contingent and experienced by others. It's this kind of network that is a really interesting part of cringe right now, because the articulation of cringe as an effect is coming after the Internet, and particularly social media, gains dominance.

I have created an Instagram account that just follows meme accounts. A lot of those are cringe TikTok meme accounts, so they’re literally scouring TikTok for the most cringe content and then presenting it. In experiencing that, I'm scrolling through hundreds of memes and videos, very rapidly modeling somebody else, and having a cringe response to their behavior. That is something that was probably not possible before this moment. If you lived in a major city, you could experience a lot of people within a local culture, however cosmopolitan it may be, but to be able to go online and see thousands of people, model their consciousness, and form an emotional response to it is a very taxing but also very rich experience of subjectivity, empathy, and the gaps between self and other. It’s this gap between self and other. This is the space that cringe articulates, but also the challenge that cringe poses. Cringe to me is an emerging form of self-awareness that hasn’t escaped the gravity-well of the unconscious. We can be aware enough to feel weird, but not aware enough to transcend the weirdness. It seems to open a big hole in subjectivity.

Identity is cringe. To me, maintaining an identity feels like a very cringe activity. I have to constantly uphold this definition of who I am that may not be true, or can be true, but is really reductive and the effort is so much work and I’m constantly kind of embarrassed by the way that it changes or the way that this identity that I’ve created demands so much from me. That’s very close to the ego and a sense of value and validity. If these start to crumble, then what are you left with? Buddhism has approached this problem of creating structures for experiencing the self without these conditions. It does so through empathy and compassion, through the Bodhisattva who vow to eliminate suffering. There's something there in terms of a teleology: what is the purpose of irony, what is the purpose of cringe? We could frame that Romantic description of irony as cope for one’s lack of infinitude. I think of cringe as cope for attachment to one’s subjectivity.

 

AM I JUST A VIBE?

Peli Grietzer: I am interested in the idea of the physical effect of cringing. In the novelette, it’s often the thing that precipitates an actual change of relationship, venue, or direction for the narrator. I love the idea of bringing back the physicality and force of the cringe into the concept and speech act of judging things in the world to be cringe. It brings out the repressed vulnerability of the judge of cringe.

The other thing I wanted to ask was if there is some kind of relationship to AI. If cringe is about awareness of the relations between the modeler and the model, these large language models in particular, and the terror associated with them for people — like, If we’re all modelable, I don’t have to keep writing books, the modeling machine can write my books for me. Anything about me that I think springs from a holistic and ineffable, unspeakable, organic whole that I'm integrated with is actually reducible, retractable patterns that the big machine in the sky can model and lay out for others to play with like a toy. I was wondering if you think that the terror of the cringe is going to have a special role to play, now that we’re worried about being literally mechanically modeled and having our models available for others to toy with?

K Allado-McDowell: That's a really interesting question. To put it another way,, “Am I just a vibe? Is my vibe cringe? What if my model does something cringe?” 

There are a few ways that AI relates to these structures, even in terms of the characters. I was thinking a lot about online subjectivity and what it would be like for Gen Z to be living through this isolation and experiencing so much of the world online. Something I observed in the language is that it's highly referential, memetic language, meme language. But it would be very difficult for people under a certain age to actually have access to all those experiences that are being referenced. It's highly referential, but also superficial. It’s not an embodied knowledge. This is something that AI also has. It's only referential. It doesn't have embodied knowledge. It only has linguistic representations of knowledge. At least, the large language models do.

Models judge. They determine appropriateness, accuracy, or likeliness to occur after a prompt, but they don’t make value judgements. In that way they can’t embody cringe.. The value judgment that we’re making with cringe is one that our body is making. We’re seeing something and reacting with this feeling of “ach.” You can't even verbalize it until after you've experienced it. It’s not something you watch and say, “That’s cringe, and now I'm going to feel cringe.” You immediately feel it.

A lot of overcoming cringe is learning what your subconscious reactions to things are. It’s impossible to cringe if you are invulnerable. To cringe is to feel what you think the other person is feeling, or at least what you would feel if you were them. That’s an entry point to empathy and expanding the range of your selfhood. If you are able to use that emotion in that way, you can unfold something. That’s where cringe can be transcendent. You're either going toward the empty subject and toward transcendence by going outside yourself, or you are stuck in cringe and collapsing into this black hole of infinitely increasing distance and judgment. I might say, “That person is cringe”, but the fact that I can articulate that makes me cringe in the process because I resonate with it, empathize with it, and feel it. That we can have a cringe experience at all points to a deeper connection between people or some kind of shared humanity. 

In terms of the terror of modeling, that's where I part ways with a lot of people in the ML space who are hard line materialist thinkers that say things like, “All this stuff is just models and you're just a bunch of models smashed together.” I get why people can think that way. “This is all an entropy minimization, and it's all a Bayesian affective and neural process that produces our subject, and we’re all just trying to keep our organism in a certain state of entropy or not.” That’s a fun way to play a certain academic game, but when people talk like that, I want to offer them a heavy dose of psychedelics and say, “Go experience something beyond your model. Experience novelty and experience things that break this closed loop of modeling and see if you can integrate that.” When an experience slips you out of your model it breaks your trust in modeling as such.