The “Creative Singularity” came and went. Now what should we do?
In 2022, text-to-image generators like Midjourney, OpenAI’s DALL-E 2, and Stable Diffusion catapulted into the public eye, prompting fervent debate about the role of so-called “generative artificial intelligence” in creative production. When OpenAI’s conversational bot ChatGPT rolled out in November, eclipsing the million-user mark in five days, concern about AI job displacement ballooned from visual artists to include authors, journalists, and copywriters. All throughout, startups and incumbents have launched a wave of new AI offerings, from voice cloning and avatar substitution to interface design and code copilots. Suddenly no category of human creative labor felt safe from automation. After months of mounting frustration about their art being included in training data without permission or remuneration, thousands of artists participated in the “No AI Art” online protest in December.
Like chess before it, art was believed to be a quintessential human endeavor, but the rapid evolution of generative AI complicated that assumption. Algorithms have been ubiquitous for years—influencing our work, social lives, and entertainment—but the cascade of innovations in generative AI has surprised even the skeptics. Concern about these technologies has also instigated renewed talk of the Singularity, a term used to describe the moment at which technological progress, generally in the form of an “artificial superintelligence,” explodes at a rate so rapid it surpasses human capability, driving unforeseeable changes to civilization and rendering a new reality in which humanity has been displaced as Earth’s dominant species (think hostile robot takeovers in sci-fi).
I don’t think the Singularity is here, and there are problems with framing machine intelligence through a reductive human lens. However, I do believe the recent advancements of generative AI have pulled us into a less apocalyptic “Creative Singularity,” upending the basic norms of creative production and associated industries.
I’ve become allergic to talk of “disruption,” but after a decade working in emerging technologies as a writer, curator, and futurist, I have to admit this doesn’t feel like business-as-usual guff. As artist-academic Mat Dryhurst recently put it, “People are understandably very exhausted by those working in technology saying everything is about to change, regrettably at the moment it does appear everything is about to change.” The advent of large multimodal models like Google’s PaLM-E, and OpenAI’s GPT-4—the latter of which researchers have claimed exhibits “sparks” of “artificial general intelligence” through its ability to problem-solve across domains without special prompting—feel like indicators of a future inflected by synthetic intelligence. At the same time, many of the zealous proclamations from prominent figures in the field feel divorced from reality. Furthermore, they bypass less extreme but more applicable lessons from the past that must be reiterated in the face of hype, which tends to erase guiding insights.
If the Creative Singularity has occurred, where does that leave us? To make sense of that, I draw from a framework I’ve developed based on philosopher of science Thomas Kuhn’s notion of a paradigm: a set of concepts, theories, or patterns that form a global organizing model with explanatory power. This framework views reality as something that evolves with humans—it’s the sum of our abilities to make consensus agreements about what is real. The technologies and symbol systems (literacy, numeracy, code, et al.) we create literally expand what reality can be. Mass-scale machine learning tools are an integral component of the contemporary paradigm, which I call “Postreality.” If, as Marshall McLuhan claimed, art operates as a “Distant Early Warning system that can always be relied on to tell the old culture what is beginning to happen to it,” the contributions of artists and other creative professionals who have been researching, experimenting, and working with AI offer crucial signals for this new world.
An Alien Lifeform
In a 1999 interview, David Bowie tries to convince journalist Jeremy Paxman that the Internet is not an incremental innovation, but a sea change in how art is created, distributed, and experienced. “I think we’re actually at the cusp of something exhilarating and terrifying,” Bowie says. “It’s not just a tool, it’s an alien lifeform.”
There is a similar sentiment today among proponents of generative AI. But what exactly is newly possible? Will there be new modes of creativity and social dynamics, and what are the impacts and byproducts of such changes? Who will be most affected? To discern signal from noise, these questions push us to assess four categories: automation and creative labor, augmentation and velocity, aesthetics and artistry, and convergence and emergence.
Automation & Creative Labor
At face value, it’s easy to appreciate that billions of people can now access powerful tools for creative expression, or that artists can automate tedious aspects of their practices. But the elephant in the room—which most immediately affects the largest number of people—is how these tools intersect with jobs and livelihood.
Mashinka Firunts Hakopian, associate professor of Technology and Social Justice at ArtCenter College of Design and author of The Institute for Other Intelligences, has researched artificial intelligence for nearly a decade, with an emphasis on its interaction with real-world systems. Institute uses speculative fiction to examine how the mythologies surrounding technologies often disguise the realities of their creation. In the book, she extends philosopher and science historian Donna Haraway’s critique of purportedly “objective” systems, which claim to represent a “view from above, from nowhere,” to AI. She highlights that all data exhibits an implicit politics, often reflective of existing power structures.
“The questions that we’re seeing crop up now around labor are continuous with the questions that have cropped up around emerging technologies and labor for years, and they reproduce many of the same blank spots and omissions,” Hakopian said in an interview. “Debates around generative adversarial networks and image generators, for example, have been grounded in the labor of the artists that’s being extracted, but fewer of them attend to the labor of the data workers who are training these models, and the labor conditions under which they’re being trained.”
The very notion of creativity as we understand it is wrapped up in historical norms, which influence the types of expression deemed valuable (and thus reproducible).
“Which forms of visuality and whose visions are being highlighted or reproduced or extracted or remixed in the outputs we’re seeing now?” Hakopian said. “There’s a strange paradox where we impute a hyper-novelty to these tools, when it is very often the case that what they are producing is ultimately a reproduction of existing canons.”
In addition to creating with generative tools, artists play a vital role in reflecting on what’s missing from them—in “misusing” them in order to discover their weak spots. Artist Minne Atairu uses machine intelligence across a range of different projects, including examinations of algorithmic beauty standards and reimaginings of Benin Bronzes in her Lumen Prize-winning series IGÚN. Such work is testament to how artists can incorporate these tools rigorously to generate novel artistic expressions and even interrogate the biases of the models used to create their work. But Hakopian cautions that we’re skipping a crucial step in seeking out artists and designers who are successfully adapting to the radical transformations underway in creative work.
“The burden of responsibility should instead be on the infrastructural layer of tech companies producing these technologies, the employers and clients who solicit labor in this art and design ecosystem, and the regulatory bodies who should be tasked with preventing the most extractivist timelines in these scenarios from materializing,” Hakopian said.
When it comes to labor, it seems AI will follow current economic and political patterns rather than supplant them. Moreover, with large incumbents, startups, and government agencies locked in an AI innovation arms race, it’s easy to see how a multipolar trap intensifies through the Creative Singularity: in seeking to decrease costs, companies consolidate their workforces, migrate creative tasks to gig (or even ‘ghost’) work, and offset ever larger populations of creatives, who then compete for the dwindling supply of openings. In fact, this is already occurring.
New jobs will also be created as a result of generative AI, and ones more nuanced than the Prompt Engineer listings making headlines. But the open questions are whether that number will keep up with the lost jobs, and how the surrounding industries and governments will respond to the shakeup.
Augmentation & Velocity
Generative AI augments our creative capabilities and the speed with which content can be produced. Alexander Reben is an artist and roboticist whose artistic research and experiments use humor and absurdism to reveal the potential and limitations of AI. In working together on both AI Am I?, his solo exhibition at the Crocker Art Museum, and our forthcoming book I Create Like the Word: Poetry in the Age of Machine Intelligence, Reben and I have been discussing what he calls “human-machine symbiosis,” a line of research he’s pursued since 2012. The term, a twist on the more conventional “human-machine collaboration,” is more than just a semantic flourish. It’s simultaneously meant to reflect his belief about the role of technology in human evolution and position his engagements with machines as expressions of emergent relationships with learning entities, rather than just inert artistic materials.
“The idea of human-machine symbiosis stems from technology as something that is inseparable from humanity,” Reben said in an interview. “Inventing stone tools and other external means of amplifying our abilities has allowed us to, for example, have more calories and the time to do things like invent science and philosophy. Technology has always been a very human thing.”
From the stretched canvas to paint pigments, all the implements of artmaking were once new technologies. Generative AI is the latest in a long line of such innovations that have expanded our artmaking capabilities. But unique to these new tools—especially buzzy new ‘AI agent,’ offerings like Auto-GPT—is the degree of agency and self-learning they have in the creation process.
“The type of automation we’re seeing now is different from what we’ve seen in other periods of automation, like the Industrial Revolution,” Reben said. “We’re now automating mental labor versus physical labor, and I don’t think we’re fully prepared for what that means.”
“I think we will see less of that kind of work because it’s less financially sustainable,” McCarthy said. “If that actually does become a trend, it would represent a huge loss in culture and our ability to process and understand the world through the art and culture around us. Because that’s what art is doing: it’s giving us a way to make sense of what’s happening. So everything’s happening faster and there’s less work that is created with the appropriate time and space to provide that.”
In 1930, famed economist John Maynard Keynes predicted that, by the early 21st century, technological progress would bring about an “age of leisure and abundance,” with 15-hour work weeks. Will the deployment of generative intelligences across general purpose tasks open up more free time for humans to explore their own creative faculties more fully? It’s a vision many would embrace. The road there will be bumpy even in the most optimistic scenarios, as generative tools will also likely create new forms of distraction alongside productivity gains, but one step in this direction is copilots, a lightweight form of AI agent that can take different roles. For creative tasks, copilots act as “blank page killers,” helping trained artists and novices alike kickstart a creative endeavor, whether that’s helping imagine openings to an essay, concepts for characters, or mockups for a series of paintings.
For non-artists, copilots might automate work to free up time for creative tasks, whether as personal assistants or as agents finetuned for specific knowledge tasks. Just as once there was “an app for that,” we can now imagine “a copilot for that,” though we’ll have to factor for hallucinations and potential alignment problems. This augmentation is a double-edged sword; where it facilitates unstructured time for some people, for others it increases competition and demands on their time (and could ultimately stratify human interaction as a “luxury” experience). But the trend of handing off knowledge tasks to copilots could foster deeper value for more human abilities: imagination, curiosity, synthesis, presence, and interconnection—all while creative capabilities are dispersed to people outside of creative professions through generative tools.
Aesthetics & Artistry
The Creative Singularity will mean shifts in aesthetics and the ways artists work. McCarthy explained that generative AI creates a new social environment that the public will look to artists to reflect on—contrary to fears that it would invalidate human efforts.
“I see the role of the artist always as having been about working with the tools of the medium available, and offering perspectives that are uniquely human or artful,” McCarthy said. “I’m not sure that can be automated.”
It might be the case that ChatGPT will be the bot that launches a billion books, but how many will interest the public enough to actually read them? And of those who do, how many would cite those as competitive with one written by a person? In this sense, the highest-touch and most conceptual forms of creativity—borne of an artist’s deep engagement with the world, their craft, and the questions animating their practice—seem somewhat insulated from automation. If anything, the public will need these interrogations more than ever to make sense of what’s going on. For such artists, generative engines would join a list of possible tools and materials that might aid in the production of a given artwork.
As AI capabilities grant more people the ability to execute professional quality creative outputs, they will continue a trend in art that has been underway for decades: centering value on the pairing of concept and aesthetic execution. Of this number, some percentage will be people who otherwise felt barred from participating in the arts; they’ll be able to produce meaningful conceptual art if they have a strong enough idea (at least theoretically). On the other hand, many more will be able to engage in new creative activities for the fun of it, rather than as a means of pursuing work or their career.
“The biggest potential is in this democratization of expression, the ability for people to create output based on their imagination, which may have been hard for them in the past, whether because of lack of skill, ability, or knowledge,” Reben said. “An obvious parallel here is the camera. Once upon a time photographic images required high degrees of skill and, you know, chemicals like cyanide to process. Now everyone has a camera in their pocket.”
In the 2010s, art created with generative adversarial networks (GANs) and other forms of machine learning had a distinct look, evident in the work of Memo Akten, Sofia Crespo, Jake Elwes, Mario Klingemann, Anna Ridler, and others. Ingrid Hoezl and Remi Marie dubbed this the “softimage” (and later the “postimage”), in which image-based works are “no longer a solid representation of a solid world but…a programmable database view.” These aesthetics have given way to more plausible and photorealistic outputs (see: the Pope in a puffy coat). But even as generative tools produce increasingly humanlike imagery, the Creative Singularity will induce new aesthetics. Writer and musician K Allado-McDowell identifies a circuit of four “side effects” of working with text-to-image engines: hallucination, hybridization, mutated language, and possession.
“Wet clay conditions the ceramicist’s gestures; AI systems sculpt the mind through subconscious ingestion of word/image maps,” McDowell writes in Side FX. “The inner world of the neural net is excavated and mimicked in the artist’s inner world model.”
These human-machine feedback loops, created as a composite of humanity (however flawed, biased, or lacking), represent a new historical context for creativity. The myth of the “lone genius” artist has long been criticized, but the Creative Singularity further fades its relevance. This also means that the models will play an outsized role in the aesthetics we encounter on a day-to-day basis—whether that’s how ChatGPT produces language, Midjourney produces imagery, or Runway interpolates video. Without the input of a wide variety of actors, this risks homogenizing creativity rather than expanding or augmenting it. Moreover, the limitations of models and datasets will also determine how visible a given medium is. For example, thus far, artistic mediums that are easily packaged for machines—text, flat imagery, and sound—have yielded more attention, investment, and innovation than have the likes of 3D/extended reality, performance, dance, and installation art. Over time, this could impact who encounters different kinds of art—and the decisions artists make in working with them. For the public, the proliferation of machine-generated content could even have profound effects on their understanding of reality.
This also evokes questions around the imitative modes of artmaking that generative AI mobilizes: covers, counterfeits, and pastiche. This is the subject of heated debate in the music industry after a growing number of AI-generated songs, recently the viral “heart on my sleeve,” featuring Drake and The Weeknd’s likenesses—and the subsequent announcement by Grimes that she would split royalties equally with any song that used an AI voice clone.
Even beyond legal or technical considerations, the ability to imitate other artists has profound implications for how artists develop to their craft. Steve Jobs famously paraphrased Faulkner (paraphrasing Stravinsky): “Good artists copy, great artists steal.” One important way artists develop their poetics, style, and distinct expressive language is by diving deeply into the work that inspires them. By treating influential works as source points, analyzing them, playing with elements, and reconfiguring them, artists develop their own proverbial muscle. When we reach a point where anyone can generate high-fidelity imitations in the blink of a prompt, the landscape shifts.
Holly Herndon and Mat Dryhurst have popularized the term “spawning,” or creating works in the likeness of others with AI. In response to “heart on my sleeve,” Herndon differentiated between being able to imitate a given artist and being able to bring an equivalent degree of care and artistic intent. In most cases, they will fall short—but things get more complicated when they don’t. Herndon writes, “The concept of sharing your identity is fascinating – someone could perform me better than me or in a context different to one I know.” And as everyone can riff in the style of all other artists, a collective culture will emerge that determines not only how young artists learn but who they encounter. The recent Wes Anderson trend on TikTok and Instagram also indicates how highly visible artists could be impacted by AI-generated riffs—compressing them into hyperbolic or stereotypical representations of their work in ways that impact their legacies.
Writer Ted Chiang, meanwhile, argues that no form of output from a large language model is a beneficial starting point for young writers: “If you’re a writer, you will write a lot of unoriginal work before you write something original. And the time and effort expended on that unoriginal work isn’t wasted…it is precisely what enables you to eventually create something original. The hours spent choosing the right word and rearranging sentences to better follow one another are what teach you how meaning is conveyed by prose.”
Puppetry is therefore another significant (and thorny) feature of both artist education and machine aesthetics, and the only thing we can know for certain is that the phenomenon will prompt major shifts in how artists develop their skills.
Convergence & Emergence
The Creative Singularity also means that creative capabilities will become commonplace among people who otherwise wouldn’t assume they could be creative. Given existing precedents—think user-generated videos on YouTube—when capabilities become available en masse, culture changes in unexpected ways, driving new mediums and forms of expression.
“Maybe our understanding of what creative output is will change,” McCarthy said. “That might end up being a different thing than someone generating an image or some text or whatever, which might become much more common and be used more in the way we communicate with memes today.”
Stephen Marche calls the coming epoch the Big Blur, because all written content will come with the question: “Person or machine?” I contend that this blur extends beyond the provenance or authenticity of the content we encounter, in fact radically altering how knowledge will be produced, organized, and applied in Postreality. As the creative impulse (however historically confined) pervades other fields, it will induce a deeper shift: best practices and insights from other fields will come to pervade each other. Creativity becomes the sort of hemoglobin for transporting ideas across domains.
Economist Noah Smith calls AI the “third magic,” referencing it as a large-scale meta-innovation that updates the ways we learn about the world, following the development of history (passing down information) and science (deducing general principles about how the world works). One way that AI—particularly deep learning—can diverge from the scientific method is its ability to recognize patterns across vast troves of data without needing any particular idea about what it’s supposed to find (i.e., a hypothesis). This approach to information means that, in a nontrivial number of instances, insights will be effective but not necessarily explainable (via the so-called “black box” problem).
“[M]any complex phenomena like language have underlying regularities that are difficult to summarize but which are still possible to generalize,” Smith writes. “If you have enough data, you can create a model (or, if you prefer, an ‘AI’) that can encode many (all?) of the fantastically complex rules of human language, and apply them to conversations that have never existed before.”
We already get glimpses of these weird possibilities in the examples of ‘AI cryptids’ Crungus and Loab, as well as DALL-E 2’s supposed secret language (i.e., Apoploe vesrreaitais).
In this way, our knowledge paradigm becomes more in-line with the workings of creativity, following Alfred North Whitehead’s claim that art “is the imposing of a pattern on experience, and our aesthetic enjoyment is recognition of the pattern.” But this “control without understanding, power without knowledge” relationship to knowledge demands increased urgency in building robust safety, ethics, and “slow AI” apparatuses, in both public and private sectors, to advocate for equitable models and development processes to ward off an intensification of the biased outcomes that we have already witnessed through algorithmic culture (i.e., predictive policing, loan evaluations). Furthermore, it demands that individuals don’t exclusively rely on AI products from major tech companies, whose market and stakeholder incentives could box in what forms of creativity can be explored in the first place.
The Creative Singularity in Postreality
In an essay written early in the Covid-19 pandemic, Elizabeth Dias outlines how apocalypse, when understood through its usage in the original Greek (apokalypsis), means an unveiling or revelation rather than the end of the world. Through this lens, singularities become critical unveiling points along a continuum, rather than single, hyperbolic cataclysms. For all of my irritation with dogmatic Singularitarianism, I do believe that the Creative Singularity is a moment of unveiling, a significant development in the evolution in human creativity. It forces us to confront the ways that creative labor has been devalued long before the advent of generative tools—and how external forces might use AI as an accelerant. It also reveals the value of human curiosity, critical thought, and analysis—which continue to elude easy automation, and which will be vital for translating what’s happening to the “old culture.” Its impacts will be felt differently by different people. For some it will entail dramatic transformations in day-to-day work, for others, it will spark new creative proclivities that might have otherwise remained dormant. For others still it won’t have much impact whatsoever.
All the while, what’s happening is that we’re building new pattern recognition engines that foster convergence of human minds and expression, forming new ecologies of knowledge and creativity. That doesn’t mean they are necessarily good ecologies—it will take work to ensure they lead us anywhere near the sunny outcomes that proponents believe are possible. In this way, the Creative Singularity is an invitation for contributors across a full spectrum of disciplines, not just in science and technology but in the humanities and elsewhere, to participate in the shaping of an emerging context for knowledge and creativity—in other words, of new realities.