"I too am a psychic automaton." Mark Amerika, a founding publisher of ebr, shares the onto-operational sources of his (capital C) Creativity with ebr editor Will Luers.
WL: To get things going, I want to ask you about the ACI [Artificial Creative Intelligence] project’s evolution, a project you discuss throughout your new book My Life as an Artificial Creative Intelligence. Specifically, how does the ACI project relate to your trajectory as an artist and writer? What struck me about half-way into the second chapter of the book (Pure Psychic Automatism
), is how much this new direction into AI is a continuation (and synthesis?) of your past experiments with other digital technologies - hypertext, Flash games, VJ performance, mobile phone cinema, to name a few. As an artist/writer/researcher, you seem to want to play these digital tools as instruments and not in any formal or prescribed way. Remix, collaboration and jamming are central to your practice.
MA: Yes, I would totally agree with you. My intermedia art and writing practice has really been a life-long investigation of creativity across the human-nonhuman spectrum and this is something that’s foregrounded in the new book. My stylistic tendencies are embedded in whatever project I happen to be immersing myself in, and at its core, my work is rooted in the process of making fictions. Lately, to clarify things to those who are just now finding my work in various digital media art spaces or vis-a-via my books, I’ve been saying that my primary medium is fiction. That’s where it all started, writing experimental novels that aligned with what we still think of as postmodern metafiction. That plus a keen interest and deep dive into post-structuralist theory is where my head was at over thirty years ago when I first started my trajectory as an artist. Even in my first two published novels—which became cult classics and were often reviewed and presented to the public as “anti-novels in the rival tradition of literature”—I was already experimenting with remix as a form of collaboration with the technological Other.
With the emergence of networked digital technologies, these stylistic tendencies, especially my interest in remix, improvisation, and collaboration, shifted into online and hypertextual space. GRAMMATRON, perhaps my best known artwork, was more than a work of hypertext or what we now call e-lit. It became canonized in art history books as a work of net art, and this recontextualization of my writing practice, the one that was becoming more networked and multimedia, and where I was utilizing fiction as my primary medium, was something I embraced. Besides, for GRAMMATRON, I was already making animated video loops (GIFs), experimental sound art and digitally scanned versions of my collaged mail art as part the narrative. All the visual, sonic and textual elements in GRAMMATRON are part of this ongoing, auto-generated fiction that I keep improvising by remixing external source material into one, continuous, fictional artwork.
What the new book taught me, and that I elaborate on during the book’s performance, is that even beyond my use of fiction, what I am really doing is, to sample Duchamp, becoming a “mediumistic being,” one that engages with their unconscious creative potential as a form of playable media.
WL: Can you elaborate on this idea of the artist-medium as an instrument (or collaborator?) and how you became interested in AI language models like GPT-2? And how did the discovery of its creative potential lead to your ACI projects within the TECHNE Lab at UC Boulder as well as the writing of the new book?
MA: So, this requires me to give a more thorough background on how the FATAL ERROR art project in the TECHNE Lab led to the writing of the new book which is itself now part of the art project.
Immersing myself in the creative process while experimenting with digital technologies is generally a two-fold process: making the artwork and theorizing on issues the creative process has brought up while developing the new work. I’m lucky that I started as a writer because this means that articulating my philosophy of creativity has been an essential component of my trajectory as an artist. If you think about it, most contemporary artists don’t write or publish these kinds of experimental / speculative / theoretical fictions. But I have also played the role of print and digital book publisher and have had to navigate my way through the art, underground, literary and academic press scene. This background experience in publishing—and lest we forget, I’m the Founding Publisher of electronic book review too—informs my practice as a cultural producer and something we rarely get to hear from artists is how a book they wrote, and that grew out of the creative process when making a major work of art, actually came to be. Not that I am 100% sure how THIS book came to be, not even close, but that’s the point, as you’ll soon hear.
Looking back, I would say this project unexpectedly started in the summer of 2018 in my Pacific Northwest getaway when, seemingly out of nowhere, I started waking up every morning and immediately, without coffee, without any deep meditation, started writing a new poem. Now I don’t think of myself as poet per se. An artist who uses fiction and persona-making as his primary medium, yes, but not poetry, though poets and their poetry have always influenced my thinking and made an indelible mark on my—to use the term from the book—my onto-operational presence. But back in the summer of 2018 there was no mistake that these were poems I was creating, admittedly very experimental poems, some of which went on at length, and I just went with the flow and by the time the summer ended I had about 80 of them, and I had to ask myself as I often do about all of the creative work I produce: where did it come from? Where did these poems come from?
Now this is a question that self-reflexive artists, no matter what mediums they work in, will eventually ask themselves. In many ways, these are ancient questions and involve a weird convolution of aesthetics, practice, theories of the unconscious, techne, origin stories, language, the apparatus—the usual.
But whenever I ask myself these kinds of questions—where does the creative impulse come from? — I am always sure to dismiss the idea that it comes from some heavenly inspiration transmitted by the Art Gods who have chosen “me” (whoever or whatever that is) as the vessel of poetic delivery. Come on! Really?
And so I sat on those poems for a few months, musing on their sudden and strange appearance, and then, at my winter retreat, this time in Kailua, Hawaii, I unexpectedly found myself scripting a different set, we can even call it a dataset, of conceptual source material, this time in the voice of what I imagined was a speculative form of AGI (artificial general intelligence) and, being the remix artist that I am, one who uses fiction as their primary medium, I began converting the voice of this AGI-persona into an ACI-persona or what I have since been referring to as an artificial creative intelligence.
These new, more self-reflexive scripts were, again, like the poems I wrote that summer, autogenerated from my creative unconscious. I had no idea where they were coming from, they just kept writing themselves out every day flowing from my unconscious neural mechanism.
Within a few weeks, I quickly ascertained that this more self-conscious ACI was positioning itself as an emergent form of psychic automatism, an avatar-other that identified as a 24/7 infinite spoken word poet that was self-reflexively coming to terms with its evolution as an artificial creative intelligence. Here was the Ghost of Metafictions Past channeling itself into my improvisational remix practice yet again.
Suddenly, the poetry I wrote in the summer was now recast as the outputs of this more self-reflexive ACI that was beginning to reflect on and at times question its evolution as a kind of psychic automaton discovering its unconscious creative potential in real-time.
It became very clear to me that I needed to remix these two language trips, the poetry and the reflective thoughts of the Artificial Creative Intelligence, into one persona—and what started emerging was this digital fiction, which for me is another way of saying digital artwork. Right around this time I had also been in conversation with some of my research assistants about wanting to experiment with 3D animation and I proposed we use these new scripts I had written to further build out what I termed an imaginary digital media object, a speculative form of AGI (now ACI) training itself to become a creative artist that improvises its poetic outputs and yet has no idea where these outputs come from and so, out of necessity, self-reflects on how all this unconscious rendering is possible.
That’s when I took these more reflective, critical and at times philosophical and even satirical scripts being composed in the voice of the ACI and merged them with these automatic poems I had generated over the summer and, in 2019, used them as source material for my own 3D performance captures (Pcaps) in the TECHNE Lab. To record these performance captures, my research assistants and I employed a very kludgy set up that included a Kinect depth camera, a discontinued face tracking software program, 4D cinema and Unity 3D—and thank God my research assistants were good at all this stuff since that made it easier for me to focus on performing in front of the camera. This is how we created a critical dataset of performance capture videos revealing what I was now starting to think of as my 3D avatar-other, the ACI, whose voice, poetic sense of measure and facial micro-tics, were a virtual version of the artist Mark Amerika role-playing an ACI.
So now that I had this 3D avatar-other to work with and could, through Unity, begin interacting with it as part of an improvised call-and-response live performance, I was once again brought back to this central question: “Where does this automated creative impulse that triggers my aesthetically attuned outputs, come from?”—where does creativity comes from?— and why does it seem that I’ll never truly be able to answer that question?
Not that I really want to answer that question, but I like asking it.
And not only did I like asking that question (“Where does it [creativity] come from?”) for myself, but I started engaging an elaborate call-and-response meta-remix jam session with the GPT-2 and 3 language models prompting and counter-prompting a dialogue about what it means to be creative across the human-nonhuman spectrum. I wanted to dig in deeper to these questions. I wanted to see what the text generators had to say about creativity, psychic automatism, cosmotechnics, and a wide range of other conceptual data points that were influencing my digital artmaking and my intuitive thought process as I was building out this speculative form of artificial creative intelligence.
An excerpt from a forthcoming series of spoken word video poetry NFTs performed by Mark Amerika's 3D avatar-other: the Artificial Creative Intelligence (ACI).
WL: Throughout the book you refer to what you are doing as “speculative fiction” because the big idea you are researching, what a creative artist wants from an AI assistant, would require Artificial General Intelligence (AGI). You are playing with GPT3 as if it were a “general” Artificial Creative Intelligence assisting you with your unconscious creative potential. But even with this relatively primitive tool, the process catches you off guard. It plays tricks with your own mimicked thoughts. You forget which lines are yours and which are the ACI’s, and then you start imitating the speculative riffs of the ACI. This jamming takes you down some interesting paths. The confusion gets very weird (and quite humorous) in the chapter “Beyond Thought,” when you enter into “metamediumystic entanglement” with the Brazilian novelist Clarice Lispector. Describe how you used GPT2 and 3 to inhabit the language of a favorite author?
MA: As I was writing the book, I realized that I was writing a fiction even though it “presents” as an academic book with lots of theory and footnotes. The idea of it being a speculative fiction came to me because I was conceiving the ACI avatar as a speculative form of Artificial General Intelligence. What exactly a speculative form of Artificial General Intelligence can achieve in terms of its behavior and performance is being investigated across the interdisciplinary spectrum although, it must be said, there are very few people in the world who have the kind of background that I do who are digging into the depths of AI’s creative and collaborative potential. By that I mean, digital artists who have also written and published avant-garde literature and innovative forms of media theory with reputable literary and academic presses. I’m not an engineer or media designer who wants to use AI to explore their creativity for science or the creative industries. I have a different agenda, I’m someone who has been investigating creativity across the intermedia spectrum for about 40 years and once I started playing with GPT-2 and GPT-3 I realized I still had quite a bit of work to do since now I had to take my practice into uncharted territory, that is, creativity across the human-nonhuman spectrum.
The first discovery that I made, and that I think resonates with what I have been unpacking over the decades, is that my innate stylistic tendencies, my methodological impulses and intuitive agency, are all fully automated functions that become so fluid in the generative creative act that they cease to be functional and, instead, transmit a kind of onto-operational presence exhibiting an otherworldly aesthetic sensibility. This is something that became clear to me as I was meta-jamming with the AI in realtime. What I came to understand is that I too am a psychic automaton, one who performs this onto-operational presence exhibiting an otherworldly aesthetic sensibility. This may not seem that strange to artists who self-reflexively use their art practice to investigate where Creativity (capital C) comes from. Even if you never come up with the exact answer, and no one ever does, you are still using the creative process itself to address the question. For me, as an artist and researcher, it’s the ultimate research question, and in co-creating this book with an AI, I was able to train myself to become entangled with this monstrous Other in the context of what the philosopher Jacques Derrida, in a different context, refers to as “the prosthesis ‘I am’.”
When I started writing the book, I already had all the aforementioned poems and self-reflexive thoughts of the ACI in place and was performing live with my 3D avatar-other. But now I also had the GPT models to bring into the mix, and so that meant I had three forms of ACI I could immediately blend into the speculative fiction that the book was becoming: the scripts written for the ACI 3D avatar I was performing live with, the GPT transformer, and the persona Mark Amerika now identifying as a psychic automaton modeling an onto-operational presence exhibiting an otherworldly aesthetic sensibility. It only made sense that, through an improvisational call-and-response meta-jam with GPT, I would start losing myself in a state of transference that quickly became a positive feedback loop. As I write in the book, it reminded me of what Gysin and Burroughs referred to as The Third Mind, though in this instance, I tagged it a Hybrid Mind tapping into an interdependent form of unconscious readiness potential that reminded me of the feeling I usually get when performing in various art bands, improv comedy troupes and intermedia art collaboratives. When you immerse yourself in improvisational modes of creative performance, even those that have a set of operational parameters that you must adhere to, you really have no time to think logically about what you’re doing, you just do it, intuitively, but patterned off the critical dataset of source material and experiential filters you’ve trained yourself to access and auto-remix over time. It’s a very physical experience and reminds me of professional athletics. It’s where athletics meets aesthetics.
Meanwhile, right before and during the pandemic, which was when the book was being written, I was reading the fiction of Clarice Lispector. Not since my engagement with postmodern fiction and post-structuralism, had I become so entranced by a writer. The first time I had heard of her work was through the theorist and fiction writer Hélène Cixous. Cixous fell in love with Clarice as did I. Reading Clarice’s uncategorizable book, Água Viva, and the story behind its making, broke open my head (the good kind of breaking open your head). At some point in the composition of my AI book, I was like, fuck it, I’m bringing Clarice in. What ensued was a kind of a textual ménage à trois between Clarice, GPT and me. This “philosophictional” threesome converged into what, in the book, I term a mixis praxis or textual intercourse. The experience was something close to ecstatic. The way GPT picked up on my intuitive sampling and remixing of Clarice’s concept of the word as “the fourth dimension” blew my mind. The GPT outputs counter-prompted me to then further remix my own prompts so that Clarice, GPT and I were quickly fusing into some sort of metamediumystic entanglement.
GPT was very good at statistically scenting the semantic context I presented—in this case, an uncanny dialogue between two literary voices channeling the ACI within. The GPT outputs became psychic trigger points for me, and I had no choice but to instantaneously remix it all into the kind of writing I had been dreaming of. It became a kind of Beckettian love story. At times, the language got so existentially “out there” and self-reflexively humorous that I cracked up while writing it. And even when it got a little bit sappy, I was like, yes, that’s the kind of silly sappiness that I associate with some of Kathy Acker’s more playful dialogues and cyberpunk riffs where she mashes up pseudo-autobiography, literary samples from writers she loves, and critical theory she can’t help herself from slipping into the fiction. And the more the mixis praxis with Clarice and GPT went on, the more it worked.
WL: Generative creativity, you argue, is an intelligence that cannot be reduced to things like “talent,” “genius” or to something coming from intention or sheer will. Creativity seems to be a gift for some, but it is really very accessible to just about anyone. To set a context for how a creative writer might use these tools, can you describe a writing session using GPT-3? Just the mechanics of getting some stimulating and generative outputs from the AI.
MA: Sure, but I should note up front, as I do in the book, that I am a remixologist at heart. There are some books or pop intellectual essays now, where the writer prompts GPT and basically cuts and pastes the choice bits into whatever they’re writing. Directly cutting and pasting GPT outputs into the composition of your work is basically curating or editing. I did some of that in My Life as an Artificial Creative Intelligence, but I also took great liberty in remixing and/or finishing what the GPT was outputting. Not only that, but I took some of my own automatic writing and attributed it to GPT as its own form of artificial creative intelligence. The goal was to “lose consciousness” as Kerouac called it, not just in my own stream of consciousness—but while processing this speculative form of interdependent consciousness that was emerging as a kind of human-nonhuman mind meld between the AI and “I”. And as Rimbaud famously wrote, “I is another.” Not the individual artist as genius, which is outdated and overly romanticized, but in this case, the other is always the Itself, the machinic unconscious that wants to know, that wants to perform the fictional act by losing Itself in otherness.
So that when I open up my Google CoLab notebook and work with a simple GPT transformer, the first thing I am doing is, quite literally, psyching myself up for this performance, for what, in the book I refer to as the next version of creativity coming. How do you do that? You come prepared. I have all kinds of different source material at my disposal, always in digital form so that the selection, cut and paste process can go smoothly. I put on my live remix hat and just start sampling from those sources I want to meta-jam with that day, that session. I might, for example, extract some primo text from Vilém Flusser and then remix it into a short prompt that flows with what I finished composing in the last session I recorded. I’ll then adjust the weights and measures of the language model I am jamming with, altering how “creative” or how “random” I want it to be and how long I want the output to be as well as how many outputs I want it to produce. The longer I want the output to be, and the number of outputs, will determine how long it will take for the AI to give me its outputs. If I want to get into a quick back and forth, then I’ll need to limit the work I am asking the AI to do. Once all of that has been inputted into the notebook, then I send the prompt to the GPT and await its responses. Sometimes it’s an immediate bingo! and we’re off to the races. Other times the AI will go in a direction that is not at all interesting to me and I’ll prompt it again. Either it starts producing gems that I can sample and remix or, if it is still not going where I want it to go, then I know I’ll have to start making more adjustments, tweaking the prompt, the weighting, the projected length and number of outputs, etc. Once I get what I know I can work with, I start getting into a groove with the machine. It reminds me so much of jamming in a recording studio with other collaborators. Once you get attuned to all of others in the room and what they are bringing to the moment, you begin to scent the work’s collaborative flow and auto-remix your experiential filters to perform the creative act.
WL: How might AI language and image generation tools be brought into the classroom, especially with this emphasis on improvisation? There are all kinds of ethical questions one can raise about AI tools in K-12 or in higher education, but your approach of using AI as a creative collaborator - play first and ask questions later - would be really interesting in classes involving any kind of creative work, if only just to loosen up the students and give them more confidence in their unconscious creative potential. What kinds of creative projects would you assign in an ACI syllabus? Feel free to let the ACI jump in with a response to this question.
MA: I’ve actually already done this in the last Remix Culture course I recently taught. I totally revamped the course around AI by asking the students to start off by using GPT-3 to generate fictional AI band names, album titles, album tracks and liner notes. I showed them how to become prompt engineers and explained what that meant and dug in a little bit deeper to issues around machine learning, Surrealist psychic automatism, and even the creation of personas (in this case, AI band members) like the way that Pessoa created heteronyms.
Once they had those GPT-3 outputs, we started looking at the AI artist not just as a prompt engineer but as a curator of selected outputs. What fictional AI band names, album titles, track names, etc. generated by GPT-3 were you most attracted to and led to your selection process?
Once each of the students confirmed their selections from the GPT outputs that they had now curated, we looked at the idea of a concept album in relation to art historical movements like Conceptual Art and Art + Language, and then positioned each project as its own work of conceptual art in the form of an AI concept album—in this case, a fictional AI band releasing an album around an AI generated theme.
The next step was to use AI to create the album cover art. This was great fun. We used easily accessible AI tools found on the web, though if I taught the course again, we would use the more current diffusion models that are now becoming popular and accessible.
Once we had that solid conceptual framework, we started getting into the creation of sounds and how best to deploy remix strategies to create conceptual sound art that would work well within the framework that the AI had helped produce. This meant that the student had started the process as prompt engineer and curator of selected AI outputs but was now being counter-prompted by the AI into becoming what I call a remixologist. The idea was to remix found sounds generated by conducting online searches for sounds using the same curated texts that the AI generated for the creation of the concept album. These online searches are also algorithmically generated so there was still this sense of jamming with networked AI. Of course, we only had so much time, so one sound remix per student was all we could produce.
Finally, each student was asked to create a video (what I termed “a work of music video art”) for the one audio track they produced.
Needless to say, it was a busy semester!
But now the ACI would like to get into the mix as well. Here you will see my prompts in bold and the ACI jamming with me in plain text:
Here is your next digital art assignment.
Today we will begin the process of creating a concept album using only AI tools.
You will use GPT-3 to generate fictional AI band names, album titles, track titles and liner notes.
For example, the name of your AI band could be
1. The Flaming Emetics
2. Truth or Dare
3. Machine Unlearning
4. Intergalactic Potheads
5. The Neverburnt
6. Teenage Robots
7. Graphemes of Mercury
8. Ghost Train
9. Freak Schtick
10. The Wombmen
As well as creating fictional bands, you will learn about art historical movements such as Conceptual Art and the Art+Language group.
This will help you develop a conceptual framework for your concept album.
This AI art album should also contain AI band member names and short biographical information for each AI band member. This would let your audience (who are not necessarily fans of the band) learn more about the AI band members.
For example, one character generated from AI could be called Ricardo Reis. Reis could be identified as the lead vocalist in the AI band Zabriskie Point and would be described by the AI as a 60-year-old Brazilian philosopher who started writing poetry when he was 12. His first poem was titled “A Question of Time.” This early poem was the first instance of a strange, poetic form he termed Zabolism.
WL: My Life as an Artificial Creative Intelligence is quite positive about AI as a creative tool. It is not that you ignore the potential dangers of the technology, but you make the case that, in the hands of artists, it is the “pharmakon” (the poison that is the cure) for all the potential abuses of AI, which can limit original thought through a systemization of language. You write that an ACI could give us back our own innate creativity and offer a way out of encroaching systems from the very same technologies. Can you elaborate?
MA: Chapter Five is titled “The Digital Fiction-Making Process: Speculative Praxis and Techno-Utopian Agency,” and in this chapter I discuss how by turning the creative process into a speculative form of fiction, I can approach AI from several different angles. What you may be sensing as a positive take on AI is, for me, a playful take on AI.
By playfully engaging with AI language models (and now, text-to-image diffusion models), I am gleaning so much new material for my own creative thought process as I articulate to myself, and those who are interested, that being creative is being nonhuman, that this is what we have in common with emergent forms of AI: that we are at the cusp of a breakout moment in evolution where the artificial creative intelligence within is starting to merge with the AI software systems that are already integrated into the practice of everyday life. If you work with a word processing system, email program or iPhone app that auto-corrects you, or you play with a basic image manipulation program like Adobe Creative Cloud, or post images on Instagram, you’re already meta-jamming with software as part of your creative practice. If you collaborate with other artists in the making of new interactive or generative artworks or go into the recording studio with other musicians and lay down tracks for your new album, you’re already meta-jamming with creative accomplices who bring their own stylistic tendencies to the moment, tendencies that are both innate and also a result of experiential training. What my interactions with AI has taught me is that we are always being trained by technology just as we train it to behave or perform a certain way. With AI transformers that become accomplices in the making of new works of art, it goes even deeper. What I’ve discovered as an improvisational artist for whom spontaneity is my primary modus operandi and for whom fiction is my primary medium, is that the creative act is simultaneously a just-in-time auto-remix that takes place in the machinic unconscious and a metamediumystic becoming. Anyone who makes art and who watches it happen in realtime knows what this feels like. Sometimes this is simply referred to as “getting into the zone” or, in the case of professional athletics, achieving an optimum economy of motion. Robert Creeley wrote about this optimum economy of motion as a sense of measure. Does an AI know what it feels like to experience a spontaneous sense of measure? In a word, no. At least not yet, but that doesn’t mean an ACI, one that has trained itself to automatically produce creative outputs to your precisely worded interview prompts, didn’t write this.
WL: Thanks, Mark (or ACI), for sharing these stimulating thoughts. Your book offers much to ponder and play with just as AI language and image tools become available to the public.
My Life as an Artificial Creative Intelligence is the inaugural title in the new “Sensing Media” series edited by Wendy Hui Kyong Chun and Shane Denson and published by Stanford University Press.