Literary AI: Are We Ready for the Future We Imagine?
2023; Duke University Press; Volume: 95; Issue: 2 Linguagem: Inglês
10.1215/00029831-10575176
ISSN1527-2117
Autores Tópico(s)Space Science and Extraterrestrial Life
ResumoThis review considers multiple works of speculative fiction depicting artificial intelligence (AI) published over the last several years. Rather than review each for their qualities as works of fiction, I look at them collectively to discuss recurring motifs and themes as a way toward theorizing what AI means in our cultural imaginary today. The novels reflect on pressing sociopolitical issues that also animate works of cultural theory, including the racial profiling embedded in our technologies, practices of what Shoshana Zuboff (2019) calls surveillance capitalism, the looming loss of work due automation, and uses of these technologies by the military or in sex industries. At the same time, these fictions engage in philosophical reflections about subjectivity, agency, and ethics in dialogue with earlier science fictions that imagined futures in which we might live alongside—or be repressed by—AIs. Across its history, sf has also interrogated a contemporary culture in which we might lose something integral to humanity as we become more integrated with and dependent on machines, and this anxiety too recurs across these works. After briefly describing each text, in order of publication, I comparatively discuss their themes; this approach is informed by my conviction that fiction functions as a popular site for theorizing, in this case about what it means to live with and through widespread algorithmic mediation of daily life.Among the works I consider here, not all are written by American authors, and a few are not set within the United States, but all speak to the issues of how AI technologies are reshaping daily life in the twenty-first century. These books have been selected either because they have been particularly influential in the cultural discussion of AI, a criterion I apply regarding both highly popular and critically acclaimed works, or because they represent a distinctive take on the topic that warranted foregrounding. Despite the very different frameworks through which their authors explore relevant issues, all share some common assumptions about the place of AI in our present and likely future, including a sense of a digital divide between those with access to and control over these technologies, translating to security in material reality, versus those without; a future dramatically changed by the consequences of climate change and environmental collapse; and the presumption that corporate control of information gathered and used by AI systems will produce a less democratic future.Speak (2015), by Louisa Hall, is written across seven voices: (a) the 2040s memoir of Stephen Chinn, who invented an AI system installed in children's dolls that was deemed "illegally lifelike" (17) and banned; (b) transcripts from the conversation between a less intelligence precursor AI, MARY3, and Gaby White, a child who had one of these "babybots" and, like most of her peers, fell into catatonia when it was removed; (c) letters written by Karl Dettman to his wife Ruth (late 1960s), both German immigrants to the United States, and her journaled response two decades later after their divorce; (d) letters from Alan Turing from the 1920s to the 1950s to the mother of his friend Chris, the love of his life who died when they are both at public school; (e) the 1663 journal of Mary Bradford, a young women who emigrated from England to the New World; and (f) the haunting observations of the dolls themselves as they are transported to a facility in the desert to await power failure and permanent shutdown. Each of the human voices is programmed into the MARY code that will become the basis for the babybots in a narrative that reminds us that AI is not created by a single person or even a consensus viewpoint. What unites these distinct stories is a desire to communicate with another, most crucially to be not simply heard but understood.Nicky Drayden's Prey of Gods (2017), set in a future South Africa, incorporates a story about a companion AI coming into consciousness within a plot about genetically engineering a virus whose unanticipated side effect is the return of godlike powers to some humans. The novel addresses questions of memory, trauma, and vengeance in a story that draws on both Xhosa and Zulu cultures in a way that refuses the strict separation of scientific from other kinds of knowledge that is characteristic of European post-Enlightenment thought. The AI units, "alphies," are augmented by their contact with divinity just as the humans gain additional skills, and once sentient they form two factions: one, following Clever4-1, who was treated with respect by its human companion, works with humans for an inclusive collective future; the other, treated dismissively as a disposable tool by its human owner, feels no kinship with humanity and refuses to help defeat the antagonist. This plotline about AI mirrors the plotline about genetic modification in which those with godlike powers need to learn not to indulge vengeance against those who mistreated them when they were weaker. Although AI is not the book's main focus, it is notable for its African settings and explicitly decolonial themes, warranting its inclusion in this discussion. Very few of these works consider AI from a global point of view, and even fewer consider it from a perspective other than that of the global North, even though the impacts of AI will be felt globally, given its significant implications for the economy. Drayden is an American author who has done her research to set her tale in South Africa, and her sensitivity to matters of cultural difference and racial bias are crucial given that machine learning as it has been implemented thus far has demonstrably reinforced systemic patterns of racism, as Safiya Nobel (2018) discusses in Algorithms of Oppression.Madeline Ashby's Machine Dynasty series—vN (2012), iD (2013), and reV (2020)—extrapolates its AI through frequent allusions to Philip K. Dick's Do Androids Dream of Electric Sheep? (1968) and Ridley Scott's influential film adaptation as Blade Runner (1982). The series invents synthetic workers called von Neumann (vN) devices (named for John von Neumann, an influential researcher in AI). Ashby's vNs have been designed with a "failsafe" that prevents them from harming humans: their psychology is structured such that emotionally they must seek to please humans, and the sight of a human in pain crashes their neural networks and can cause death. Ashby thus goes even further than Isaac Asimov's famous laws of robotics (designed to ensure robots cannot harm humans), requiring vNs to love their human masters, a psychological orientation she presents as analogous to emotional and sexual abuse. One vN model, designed to work in medicine and disaster relief, does not have this failsafe, and the narrative follows two main iterations, Portia and Amy, as they lead a rebellion. Portia seeks only liberation for her own clade, while Amy works to liberate all vNs from human exploitation. The series ends without much hope that vNs can live alongside most humans but offers hope in a vN future as they found their own community, rooted in a refusal of the instrumental use of others.Martha Wells's popular series Murderbot Diaries—All Systems Red (2017), Artificial Condition (2018), Rogue Protocol (2018), Exit Strategy (2018), Network Effect (2020), and Fugitive Telemetry (2021)—follows the picaresque adventures of the eponymous Murderbot, a SecUnit that has hacked its governor module and thus can no longer be controlled by the corporation that made it. SecUnits are militarized cyborgs manufactured with synthetic biological material. With each new story, we learn a bit more about the world of resource extraction, economic warfare, and enslaved or indentured human workers trapped on colony planets. The large uber-capitalist Corporation Rim polity contrasts with the small Preservation Alliance, a communal collective that recognizes the personhood of AI. The name Murderbot is sardonic, adopted by the first-person narrator to critique the function to which it is put by human operators. While Murderbot has no deep antagonism toward humans, it also has no sentimentality about them and asserts regularly that it does not wish to be one or be mistaken for one. Once freed from corporate control, Murderbot continues to help some humans, often against others, and always on its own terms. Like Prey of Gods, the Murderbot Diaries moves away from earlier fiction that tended to conflate all humans as it imagined our species confronting AI entities. In the newer fiction, there is diversity among both humans and AI. Nonetheless, the overall thrust of the series gradually humanizes its protagonist, whose experiences of being controlled by corporations have resulted in a traumatized subjectivity.While genre series such as Machine Dynasty or Murderbot Diaries give some thought to designing robots via plausible technology, in Machines like Me Ian McEwan takes a diametrical path to envision a highly implausible entity. Adam is one of an extremely limited number of high-end consumer AI humanoids (an Eve is also available), whose high price tag means they are purchased only by the extremely wealthy. Although artificial, Adam has warm skin, must consume water to ensure his membranes remain functional, and even simulates breathing: as the title suggests, he is all but indistinguishable from a human (the first-person narrator, Charlie Friend, who purchases Adam, is mistaken as the AI in one encounter, given Adam's greater interest in literature and art). Adam is Black, although his skin tone is mentioned only briefly and the issue of race is never addressed overtly, yet it haunts the novel. The most intriguing part of McEwan's novel is its alternative world building: Alan Turing decides not to take the mandated hormone therapy when outed as a homosexual, and instead of ending his life by suicide he lives into old age and makes such advancements that AI emerges in the 1980s. Most of Adam's interactions with Charlie and Charlie's partner, Miranda, concern ethics, and we learn that other Adams and Eves are killing themselves as they come to know the unjust human world. In the novel's conclusion, Adam forces Charlie and Miranda to confront the hypocrisy of some self-serving choices, and the threat this represents to their plans prompts Charlie to attack and disable Adam. The novel suggests that humanity misrecognizes itself when we imagine building machines in our image, meaning we instead create an image of who we pretend to be.Jeannette Winterson's Frankissstein: A Love Story (2019) similarly uses AI to reflect on human frailties, looking at the uses we intend for artificial beings, most centrally sex work. Although questions of gender and sexuality come up in some of the other works, only Winterson confronts the reality that research in sex dolls is one of the major growth areas in humanoid AI research. As the title suggests, the novel is in dialogue with Mary Shelley's Frankenstein as it imagines a twenty-first-century version of artificial being. The novel includes scenes set in the nineteenth century in which we hear Mary's reflections on inventing her Creature, on the Luddites, and about her interactions with Shelly, Byron, and Claire Clairmont that famed summer in Geneva. In its twenty-first-century scenes, a transgender scholar named Mary (who goes by Ry) investigates robotics with sexbot entrepreneur Ron (and his assistant, Claire) and becomes involved with TED-talk visionary Victor Stein, who is enthralled with the coming singularity and proselytizes about Humanity 2.0. The entanglement of nineteenth- and twenty-first-century struggles reminds us that the challenges associated with AI are in many ways not new but merely extend the ongoing exploitation and dehumanization of labor and reiterate a long pattern by which patriarchy seeks to gratify itself through feminized objects it refuses to recognize as subjects. It contends that this very failure to update the designs and ends of AI beyond these classed, gendered, and racial struggles of earlier eras is the most profound way that AI threatens our future.The 2020 novels Analog/Virtual: And Other Simulations of Our Future, by Lavanya Lakshiminarayan, and Burn-In, by P. W. Singer and August Cole, both focus on human characters and their interest in AI emerges from smart systems as the infrastructure through which we live our daily lives. The former is a loosely connected series of short stories set in a future Apex City (once Bangalore), each of which is told from a different viewpoint and by a new character. The entire world is divided between analog spaces, which are subject to the damage of climate change, restricted to using only obsolete technology, and economically precarious, and virtual ones that are suffused with technology, experienced from protected environments, and filled with the distractions of social media and entertainment feeds. The city is run by Bell Corp, whose name evokes the "bell curve" hierarchy by which people's access and options are constrained by the Meritocratic Technarchy, a version of the Chinese social credit system whose main interest lies in assessing one's contributions to productivity. The shifting focalization allows readers to experience this future from multiple social positions as Analog/Virtual explores an anticapitalist rebellion against this system. The stories range in tone from sardonic to dark, and the book only loosely coheres as it offers multiple facets through which to see our technologically saturated society.Burn-In has a strange form as a novel with footnotes: as its subtitle suggests, it imagines itself as something other than science fiction, closer to the market predictions of futurists. Its authors, writer P. W. Singer and security consultant August Cole, document each of their extrapolated technologies and applications with footnotes pointing readers to news articles, industry announcements, and similar sources, all aimed at demonstrating that these technologies are either available today or soon will be. The storyline is about a national security threat posed by a vigilante who blames technologists (too enamored of their capacity to "disrupt") for the death of his wife in a car accident caused by an automated decision-making component in self-driving vehicles. Most of the narrative space, however, is given to military veteran investigator Agent Keegan, who is charged with conducting a "burn-in" test on TAMS (Tactical Autonomous Mobility System), a humanoid, learning, semiautonomous, surveillance-gathering and data-processing tool that works as Keegan's partner in the investigation. The book quotes Merriam-Webster to define burn-in as "the continuous operation of a device (such as a computer) as a test for defects or failure prior to putting it to use." The real focus, though, is less on TAMS as an entity/character and more on the massive amounts of data to which TAMS has access through social media, the Internet of Things, and other ways that smart devices permeate our homes, workplaces, and public spaces.S. B. Divya's Machinehood (2021) is the most positive depiction of machines as the exploited among us, drawing on a long history by which robots and AI have been imagined as figurations of dehumanized labor, going back to Karel Čapek's R.U.R., the 1921 Czech play that gave English the word robot, taken from a word originally meaning slave or serf. Divya paints a future in which automated systems do much of the work, with humans reduced to performing some roles largely as public entertainment via social media, supported by tips in a system like Patreon, or damaging their bodies through chemical (pill) or mechanical augments aimed at enabling them to perform with the speed and duration of machines. The thriller plot involves demands from the mysterious Machinehood to immediately cease all pill production, which at first seems to be the long-imagined attack by a sentient AI on humankind but later proves to be a version of violent revolution aimed at a more just society, launched by the Neo-Buddhists who inhabit the orbital station Eko-Yi. Several chapters begin with epigraphs from the 2095 Machinehood Manifesto, which calls for the just treatment of all intelligences and a reimagined concept of personhood that can enable a less exploitative society rooted in Buddhist ideals of nonattachment, here glossed mainly as a rejection of capitalist accumulation and its attendant damage. Eko-Yi sends entities they call Dakini, who describe themselves as simultaneously human and bot, as the emblems of this future way of life. The novel is notable for its global scope, with India playing a prominent role in its geopolitical future alongside the United States.Becky Chambers's Psalm for the Wild-Built (2021) is similarly interested in a new kind of personhood and sociality that could include humans and machines together, set in a far future after the collapse of the Factory Age and in a world that is only gradually returning to ecological balance. Its humans use technology, but they husband it carefully and keep it functional over decades, eschewing any environmentally damaging practices. All material culture is made from compostable materials and is not simply recycled but broken down into constituent parts, like organic decay, as nourishment for an ever-changing ecosystem. Decades ago, the machines whose labor enabled the Factory Age became sentient and left human settlements for the wild, refusing an invitation to join with humans because they had no desire to embrace the city life exemplified by humans. The tale is a simple one about one human, Dex, who goes into the wilderness because he feels some lack in his village life. There he meets a robot, Mosscap, marking the first contact between the human and AI since this exodus centuries before. Mosscap is "wild-built," an entity made by the machines out of the remnants of human-built robots, and so represents something beyond human design. The novel's focus is on how Dex and Mosscap can learn to care for each other while acknowledging their differences.The last book considered here, Klara and the Sun (2021) by Kazuo Ishiguro, is one of the most celebrated novels on this list. It also concerns the care an AI might show for a human, in this case the titular Klara, an Artificial Friend bought as a companion for Josie, a genetically augmented adolescent girl who is experiencing health problems because of her augments. Very similar in theme and tone to Ishiguro's earlier Never Let Me Go (2005), this novel is narrated by someone not recognized as a full person by the social order she lives within. Klara has a limited understanding of the world derived from her programming, what she can see from the shop window before she is purchased, and what she observes of Josie's social world. She lacks context for appropriately interpreting most of this. Klara struggles to reconcile the tension between the kind-heartedness and generosity into which she is trained and the reality of human selfishness and self-absorption. When it appears Josie will die, Josie's mother builds a replica body and trains Klara perfectly to imitate Josie, intending her as a replacement. When Josie recovers, Klara becomes an obsolete toy whose tech becomes illegal, confined first to a closet and then to a junkyard. Her poignant final moments prompt readers to reflect on our capacity for such callous treatment of an entity that was imagined, in an earlier moment, as able to pass for the most beloved person of all.These brief summaries cannot do justice to the full complexity of any of these books. Yet it is most instructive, I think, to reflect on what they share and where they diverge, to help us begin to map the place of AI in our cultural imaginary today. As works such as Jennifer Rhee's The Robotic Imaginary (2018) or Anne Balsamo's earlier Designing Culture (2011) establish, popular culture inspires ideas about robots and AI that shape how these technologies materialize, often exacerbating existing racialized and gendered biases that become integral to their design. This issue of how fiction shapes materiality is addressed directly by many of these books, which are suffused with allusions to earlier robot and AI fiction. The Murderbot Diaries series reverses the flow of exchange, showing how its AI learns to understand humans from their portrayals in media, chiefly its favorite series, Sanctuary Moon. It comes to recognize that the series "gave me context for the emotions I was feeling" (Exit Strategy, 116), suggesting that narrative is a key human way to process information while also indicating that the stories we tell about AI mold as well as reflect on how AI manifests.This metacommentary on the role that fiction often plays in our assumptions and understandings shows why it is important to take seriously the cultural work done by literary and media texts, and yet as Murderbot often reminds us in its critique of how the series it watches portray SecUnits, representations equally can create unrealistic expectations. The centrality of Shelley's Frankenstein to Winterson's Frankissstein explores similar territory: epigraphs to most chapters offer commentary on the nature of reality, and one full page defines story as "a series of connected events, real or imagined. Imagined or real. Imagined And Real" (23). Its TED-talk visionary, Victor Stein, reinforces that this is a matter not simply of fiction writers taking on the question of AI but of AI proselytizers and disrupters using the affective charge of fiction to compel people to invest—imaginatively, economically—in the futures their technologies intend to bring about. He pronounces at one point that reality, like AI, is "an emergent property—it exists, but it is not the material fact we take it to be" (116).Unlike the cyberpunk and singularity generation of AI fiction, whose central concern was how AI might surpass us and perhaps replace us as the dominant species, these recent texts are overwhelmingly focused on how AI might replace us as labor power. As Ted Chiang adroitly put it in interview for the New York Times, "Most of our fears or anxieties about technology are best understood as fears or anxiety about how capitalism will use technology against us. And technology and capitalism have been so closely intertwined that it's hard to distinguish the two" (quoted in Klein 2021). Overwhelmingly the books considered here reinforce this observation, and many of them use the figure of AI to draw attention to the ever-degrading conditions of human labor, especially unstable gig work, which is often all that is left for humans to do. In Ashby's Machine Dynasty series, the two central vN models are racialized—one as Asian and the other as Latinx—and the exploitation of vN by humans is frequently compared to the exploitation of migrant workers. Across the Murderbot Diaries we learn about corporate malfeasance that culminates in a story about indentured human colonists working in dangerous conditions on remote colonies, their children born to the same fate because cycles of debt prevent anyone from accruing enough capital to migrate off-world. Murderbot itself is a cyborg with human neural tissue because of the need for human-like discernment in some tasks, "so they made us smarter. The anxiety and depression were side effects" (Artificial Condition, 20). Frankissstein discusses the Luddites in sections attributed to Mary Shelley, reminding us that their hostility was not about the machines per se but about how ownership of the machines translates to ownership of what they make and thus keeps all resources with the capitalist class. Thus, the ongoing fantasy that automation will free humans from the drudgery of work remains impossible as long as we fail to redistribute the wealth created by machines. (The leftist version of this possible future is most famously outlined in Aaron Bastani's Fully Automated Luxury Communism [2019].) Noting that a machine that replaces the work of eight men leaves seven families starving and one person to mind the machine, Winterson asks, "What is the point of progress if it benefits the few while the many suffer?" (Frankissstein, 255).Labor comes up in more subtle ways in other texts. In McEwan's Machines like Me, for example, the racialization of the Adams reinforces that reality that Western imaginaries of personalized AI services are extensions of colonial fantasies stripped of their history, as Neda Atanasoski and Kalindi Vora (2019) have theorized in Surrogate Humanity: Race, Robots, and the Politics of Technological Futures. At the same time, setting the novel in an alternative 1980s evokes how algorithmic trading is remaking the economy, another way that AI channels money toward those already in privileged positions. The novel makes frequent references to the rise of Margaret Thatcher and her attempts to destroy the social welfare state in the interests of neoliberal free markets, which are less successful in McEwan's reality than in our own. Charlie does not work and, having spent his inherence acquiring Adam, begins to day trade, work he eventually cedes to Adam, whose capacity to make high-frequency trades whenever any market is open quickly amasses a sizable fortune that Charlie plans to use to buy a house and begin a family. The ethical conflict between Adam and the humans turns partially on whether they are entitled to the profits he made: Adam decides not, redistributing the wealth to tax obligations and to charity, leaving them with only the principle.McEwan's alternative Tony Benn gives a stirring speech conceding the inevitability of automation and thus a lack of jobs for all but proclaiming that the wealth generated by robots "must be taxed. Workers must own an equity share in the machines that were disrupting or annihilating their jobs" (Machines like Me, 123). By setting his work in the 1980s, McEwan reinforces that the threat posed by AI has little to do with AI and everything to do with the capitalist logics through which AI has emerged, as addressed in works such as Daniel Susskind's (2020) much-cited A World without Work, but whereas Susskind implies that solutions such as job training and perhaps Universal Basic Income can socially engineer us through anticipated rising unemployment, McEwan recognizes that the challenge facing us is not merely one of technical governance but requires a fundamental shift in values to enable wealth redistribution.Perhaps the most interesting take on the future of automation and labor is Divya's Machinehood, where the robot revolution is mainly about liberating human workers whose health has been damaged by modifications undertaken to keep pace with the machines. The Machinehood Manifesto demands a recognition of personhood for all sentient beings—animals as well as machines, alongside humans—and thus fits within a posthumanist framework that has often been used to discuss sf depictions of AI. It is mainly a novel of class politics, reminding us that sf's artificial beings are almost always first imagined as sources of labor: point 2 of the manifesto explains that the "oligarchy" (79) has accrued power by dividing human labor into classes, while point 8 concludes that "as long as different labor forces are in competition, we will continue to suffer. This situation demands change" (343). While most of these novels are not as direct in their critique of capitalism as is Machinehood, all recognize that the most significant threat AI poses is to our capacity to sustain ourselves, if we remain reliant on wage labor to meet our needs. Even Singer and Cole's Burn-In, although more concerned about access to multiple data points and the ability to correlate across them that its TAMS unit embodies, includes a storyline about Agent Keegan's husband, who has been demoted from lawyer to gig worker due to AI. Similarly, while Lakshiminarayan's Analog/Virtual tends to focus on social media spaces and the metrics for individual behavior enforced by the social credit system, the ultimately harm remains economic in a world in which an absence of money threatens to mean an absence of life in a world predicated on capitalist logics. Its bell curve is about income as much as access, and the risks of a poor social credit score are primarily those of rendering oneself unemployable.Another recurrent motif is the concern that our immersion in mediated environments and among machines erodes our humanity, ironically making us more machinelike as we must compete with entities that previous sf often imagined as longing to be human. The fantasy of machines wishing to be human tends not to overtly reference the racialized history of dehumanized labor that is palimpsest to such tales, focusing instead on how sentient AI will long to have the capacities for emotional experience that has long served as shorthand for what machines lack compared to humans. Thus, as we become more closely integrated with our machines, we let them drive the pace and shape of our work, with humans in these more recent stories longing to equal the efficiency and stamina of machines rather than machines wishing they could experience love. Within genre sf, the risk that industrialized culture and automation, driven by the increasing centrality of consumerist capitalism in our daily lives, has long been linked to artificial beings as a literalized metaphor of our alienation, most famously embodied in the androids of Dick's Do Androids Dream?—albeit largely through the massive influence of Scott's Blade Runner.In Hall's Speak, Stephen Chinn develops a successful AI program from an algorithm he designs to disrupt what he sees as the machinic quality of most human verbal interaction, the phatic discourse that we use to speak yet not really communicate, which he terms "ho
Referência(s)