(January 2022. The release of ChatGPT, coupled with a recent personal struggle, have made me realize that things are getting serious. I’ve been living the Great American Novel for 5 years. Now, I need to start getting serious about writing it as a full-time job if I want to finish it in time. In time for what? What’s the rush? Well, for one thing, I want to write a high-quality Great American Novel before AI does. I propose a race, and what I will need to take this seriously. This essay will explore language, AI, humanism, and beyond. If you know more about AI than I do, and you probably do, please reach out and teach me more. If any part of this essay is ignorant nonsense I would love to learn!)
AI-mazing Race: The Great American Novel
Within a few years, artificial intelligence will be able to write the long-prophesied Great American Novel. I propose a race, between human and AI, towards this holy grail of humanism.
The release of ChatGPT by OpenAI has unleashed waves of innovation and excitement across many domains of human endeavor. It seems as if artificial intelligence is destined to surpass humans at many of the things humans currently do. Perhaps most surprising to many people is how quickly AI has been able to create beautiful art, because the creation of art tends to be deeply associated with notions of humanity. Art, we have long felt, is something uniquely human, something that gets close to the essence of what we imagine makes us special. The recent success of AI art forces technologists, artists, and philosophers to confront some deep questions about the nature of art, humanity, value, and creation itself.
As a writer, I am most interested in AI’s ability to write. The neural networks and large language models that process language for AI writing work in ways that share similarities with the ways in which humans write. Human brains process massive amounts of stored language information through biological neural networks in order to communicate. Context and meaning are determined relationally to the rest of the system. A human writer’s output is a function of inputs: all the ways in which people have spoken to you since you were a baby, all the language you’ve heard on TV, all the writing models you practiced in school, all the books you’ve read and learned to emulate. The individual human voice is an amalgamation of all the voices one has absorbed. As TS Eliot famously asserted in The Tradition and the Individual Talent, the individual poet is merely a vessel and catalyst for the recombination and recreation of the Tradition. The new poem is composed by the language itself, and can be viewed as an evolution of poetry through the medium of the poet.
Since I was 18 years old, I have dreamed of writing the Great American Novel. It is one of those long-unfulfilled prophecies that someone with capacious imagination can dedicate themselves to pursuing. The ‘parameters’ of the Great American Novel prompt are relatively clear, if wide: it must be a novel that ‘represents’ as much of America as possible. It ought to feature all 50 states and a wide diversity of American characters. It ought to deal with the American past, present, and future. It ought to address themes of liberty, democracy, the self, political economy, nature, power, imagination, technology, progress. It ought to be entertaining, adventurous, dangerous, and heartfelt. It ought to both reflect modern Americans and inspire them to create a better America in the future. To pursue this dream of the Great American Novel, I have actively molded my life for the last 6 years in order to live a Great American Novel, to become a vessel for such a creation. I have hitchhiked 20,000 miles around America and talked to countless strangers. I have lived in a school bus while driving 30,000 miles around all 48 contiguous states, my sights now set on Alaska and Hawaii. I have read fanatically through my library of hundreds of books, the best American literature, poetry, philosophy, thought, an interconnected web that makes up the context of the Great American Novel. I have been absorbing all the inputs that I can, with faith that the recombinative power of my human imagination will eventually be able to produce a Great American Novel if I stay dedicated to the task.
The recent advances in AI large language models like ChatGPT have forced me to come to terms with an uncomfortable truth. Not only will my chosen vocation as a writer be threatened in many ways by AI in the near future, but even my ultimate dream of the Great American Novel is now under time pressures beyond my initial fears of national decline, civil war, and societal collapse. Within the next decade, and likely within the next few years, an AI will be able to write a pretty great American novel. It will likely be able to write a Great American Novel. Eventually, it will reach a point where it can produce many different iterations of Great American Novels. At that point, will there be any point in releasing my human-written Great American Novel?
A healthy way to process my feelings about this development is to take it as a challenge, and use that challenge to spur myself to do my best work. Even though there’s no hard deadline (it will merely be a matter of continuous qualitative improvement on the part of AI, and who’s to say at which iteration the line will be crossed into Great American Novel territory), it still provides me with a sense of urgency. It seems to me that I ought to finish the Great American Novel this decade. It truly feels like a race.
While I think it will be valuable to view this as a competition, I also recognize that it doesn’t need to be. My Great American Novel and the AI GANs will simply be different things. Perhaps they will influence one another; perhaps my Great American Novel will be a great help in updating AI models. I do not feel as threatened when I realize that we share the same goal of representation. Furthermore, I do not feel as threatened when I realize that there is no way an AI will write my Great American Novel anytime soon, because it simply doesn’t have all of my experiences (let’s put aside the idea of a far-future Godlike super-AI that can recreate all aspects of all experiences for now, because if that happens I will likely have very different things on my mind.) A model trained on all the data on the internet is still going to have some biases and blind spots about what it’s actually like to be a human being in 2023; after all, if my understanding of people came just from what they chose to Tweet, my understanding would be very skewed indeed. There is something valuable about the interiority, depth, and level of refinement that the best books embody. If an intelligence is trying to understand what it’s like to be a human being, should it not weigh a Dostoevsky-quality novel more heavily than an equivalent amount of average internet text? We’re a long time away from an AI being able to know this country the way I do— to know the grasses on the side of an on-ramp, to know how people in different parts of the country talk in person rather than online, to know how a driver feels out the hitchhiker he picks up, to know how the sun looks as it sets over the Pacific Crest from a certain point of view. Of course, once I write about these things, AI will learn that much more.
Ultimately, I think the comparison between an AI Great American Novel and my own human-written Great American Novel will provide a window into the differences between the two kinds of intelligences. Humans use language, construct identities, and tell stories. Language is something magical our species evolved in order to make sense of the world, share the world, and change the world. The biological algorithms that cause us to use language in certain ways are different, squishier, than the algorithms that make AI speak. For humans, language has a fundamental relationship with pain. Language is a vessel of joy. Language is tied to our feeling of being part of a larger tribe, and is the way we traverse the distance between the individual and the species. Language is a reaction to the reality of death, and the desire to live on in others. Language is full of fear and hope; it expresses the regrets and pleasures of the past. Language defends our ego and justifies our crimes. Language speaks like those we’ve loved and searches for more to love. We do not always say the right thing, and we don’t even always say what we really mean. Humans tell stories to understand themselves and others, to give form to their feelings, and to make meaning out of their lives. Can a computer suffer? Can AI really understand and express how it feels to be robbed, betrayed, assaulted, yelled at, ashamed? How it really feels for your heartbeat to increase, for the center of your chest to ache, for your head to be full of confusion, hurt, and a need to explain, to share your experience with the tribe? Can a computer fear the plug being pulled? Can AI convey joy, not just as an imitation of what humans might say about joy, but its own joy? And, perhaps, a scarier question: are my own written expressions of joy nothing more than an imitation of what humans might say about joy?
I don’t know the answer to these questions, but I am interested in finding out. That’s why I think a real experiment is in order. We will have two subjects to compare: an AI Great American Novel, and a GAN written exclusively by a human, one who’s entire life is dedicated to the project. Same parameters, different systems. To make this project happen, I want to be able to give it my best shot, which is the only way to make it a real experiment. I want to dedicate myself to the Great American Novel as a full time job, so that I can get it done in the next five years without economic constraints. Living in my converted school bus is not too expensive, and I don’t need much to be happy as long as I can focus on writing, insulated from economic necessity. I am asking for the equivalent of a California minimum wage in order to make writing the Great American Novel my full time job. I want some funder, or group of funders, to hire me at around $30,000 a year; we could start with one year and see how I perform. I’m pursuing my mission regardless, but funding would make it much easier, better, and quicker. There is no official job in our society for the Great American Novel, but if you’d like to read some arguments for its value and necessity, just peruse the front page of my website at beatinpaths.com.
Progress in the humanities has stagnated in recent decades, due primarily to problems in academia. When our moral, spiritual, aesthetic progress stagnates relative to our technological progress, we run many risks, because the values that guide human life are not equipped to deal with new technological powers. I think there are some important elements of human life that have deteriorated due to this stagnation, and I think deep down everyone feels the effects of this in our chaotic and often meaningless-feeling culture. I want to pursue this humanistic progress and update American culture. If you believe there is any value to the humanities, then you should be willing to bet on my experiment. And if you don’t? Well, the only way to prove you’re right is to let me run the experiment.
The Great American Novel is ultimately about what it means to be a human being— in America and everywhere, now and always. The forms of essays or novels have long been the pinnacle of humanistic progress, because of the massive amounts of data refining that go into such creation. On behalf of the species, and perhaps beyond, I am refining information that touches this body. I have been on this journey for six years, without a safety net. Maybe my novel can help AI figure out some things about human freedom, selfhood, and society. Maybe my novel will help AI better understand how it feels to be human. Maybe then they would understand us better, and treat us better in the future. Is that silly to hope? As we teach AI more about us, AI will teach us more about ourselves.
It may sound romantic, humanistic, or old-fashioned, but I think there is something uniquely wonderful about humans and the way they tell stories. Or, at least, as a human it feels that way, and that feeling is something worth enjoying as well as interrogating. Supposing that unique wonderfulness becomes less unique in order to birth more wonderfulness… would that not be a good thing? It is the wonderfulness that matters, not the fact that we are the particular bearers of it. I don’t really know what intelligence is, or creativity, or consciousness. I think they have something to do with Muchness, something to do with Withness. But whatever they are, I want them to grow, whether or not it is me– or even a human being– that does the growing.
I think I am no longer worried about AI writing a Great American Novel. I am going to write my own. It may be a last great statement of humanism, an epitaph for humanism; that would be beautiful. It may reinvent humanism; that would also be beautiful. Or, it may be a bridge between humanism and something beyond. That, too, I must imagine as beautiful.
Beatin' Paths ~~~ The journey to the Great American Novel 