Once upon a time, there was a tale about artificial intelligence called Snow White. In the story, the Wicked Queen has a smart mirror. When activated with the command “mirror, mirror,” an embedded voice assistant tells the Queen if she is the “fairest of them all.” Of course, today’s technology renders the mirror per se feasible with little effort. It shall have a camera, connections to other smart mirrors in the kingdom, and a metric to evaluate the Wicked Queen’s appearance against that of other users. This gadget might use a machine learning algorithm with training sets derived from People Magazine’s Most Beautiful list, or it could be based on a series of up- and down-votes. However, the apprehension with this contemporary version of the famous Grimm fairy tale as outlined is not its technical feasibility. Rather, the story reflects outdated values within a modernizing society. A truly smart mirror would tell the Wicked Queen that her obsession with triumphing using a singular beauty standard, one that prizes pale skin and youth, is misguided, reductive, and futile. “While we are on the topic though,” the mirror would say, “here are seven ways you are washing your face wrong.”
The Wicked Queen and her smart mirror is a telling analogy of the current state of our relationship with artificial intelligence: the technology has advanced to achieve astounding feats, but its value system is lingering behind. In turn, this is symptomatic of a stagnation in our own framework. Even though the social movements of the last century have introduced many nuances to complex issues such as race, gender, and power, their mirror images in AI development remain overwhelmingly simplistic, reductionist, and sometimes laughably clueless. “The way we design decision-making processes in computers is certain to replicate our own biases,” argues John Palfrey in Line Drawing Exercises: Autonomy and Automation.
The list of examples to draw from is endless. However, the argument is best illustrated through three popular narratives in contemporary AI depictions — that of the robot girlfriend, the invisible laborer, and the despotic overlord. Through these allegories emerge the reductionist ways common discourse treats the issues of gender dynamics, labor, and power structures — all incredibly nuanced and complex ideas that are undergoing revolutions of their own. Yet, when portrayals of these polemics venture into the AI realm, they reflect the precise defects in society’s complexion. In other words, these are not problems with the projected technological advancement of artificial intelligence; they are mirror images of our own flawed attitudes toward humanity that we are in turn forcing onto AI. In How To Become A Centaur,
The Robot Girlfriend
It is a truth universally acknowledged, that a single man in possession of engineering skills must be in want of a robot girlfriend.
This kind of portrayal is dangerous, and it extends beyond the scope of AI. It is a reductionist misrepresentation of #metoo that being “gross” and anti-social are necessary and sufficient conditions for committing sexual crimes. Gropers are painted as clueless about the general decorum of social interactions when in fact they are adults making deliberate choices. At the same time, similar atrocities committed by handsome men are hardly discussed, and this is especially blatant in the tradition of AI storytelling. In fact, the biggest continuity problem with Blade Runner, arguably one of the most iconic movies about artificial intelligence, is that the male lead, played by Harrison Ford, unambiguously rapes the only female character with a significant speaking role (played by Sean Young) and the incident goes by completely undiscussed thereafter. More surprisingly, we learn in the 2017 sequel, Blade Runner: 2049, that the two lived happily ever after and even produced a legendary child. The idea that a few well executed camera pans can resolve and transform indisputable assault into child- and plot-bearing love is ludicrous. It is not shocking that the scene was forced onto Young by surprise, as she admitted in Dangerous Days: Making Blade Runner. As her character Rachael submitted on screen, Young buried her tears and became a ghost of the franchise. A similarly uncomfortable dynamic emerges in the sequel between Ryan Gosling’s character K and his holographic AI girlfriend. K’s ability to love, which is emblematic of his being a more advanced replicant model, is passable yet still narrowly directed at a woman whose commercial existence is hinged on pampering to his needs. In one scene, she is seen powerlessly frozen and then exasperatedly wiped from existence. Neither of the aforementioned incidents, in two films set 30 years apart, is further remarked upon in the plotline. So, the moral lesson these stories seem to endorse is that treating a female robot like a geisha at will is scandalous, unless one looks like Harrison Ford or Ryan Gosling.
Why do we care about the fate of holograms and replicants in dystopian lore? The allure of an artificially intelligent robot girlfriend, as presented by most science fiction writers, is that she cannot refuse the male protagonists’ desires and advances (and yes, it is always a male lead); saying “no” is simply not in her program. In the rare case where a female robot, abused and aware, voices her concerns and seeks help, she is eventually silenced and dismantled, like on Silicon Valley, with her fleshy mask plopped mercilessly into an e-waste bucket. Jia Tolentino describes in her New Yorker exposition on “incels” — an amorphous community of involuntary celibates whose member is responsible for the recent Toronto vehicular attack — that the infamous group trains men to see “women in a way that presumes that women are not potential partners or worthy objects of possible affection but inconveniently sentient bodies that must be claimed through cold strategy.”
The trope of the robot girlfriend and, more importantly, her sexual predator proliferates. Comedian Dave Chappelle's Netflix Special — The Bird Revelation — jokes that if Brad Pitt did what Harvey Weinstein had done, the public’s reactions would have been different; women would have acquiesced! This obvious fallacy is not limited to incendiary comics; it bleeds into our daily lives. For instance, a recent MIT-wide sexual misconduct prevention training
This is a problematic and reductionist microcosm of how we are taught to see this issue. Other people can be bad; we ourselves cannot. In fact, one has to be so otherized to become a sexual predator, and not ever shampooing again seems to be an initiation requirement. Kevin Slavin’s adage, “you are not stuck in traffic — you are traffic,”
Grappling with the nuances and complexities of gender dynamics is difficult and requires comfort with the unknown. In Resisting Reduction,
The trope of the robot girlfriend is not a tale about AI; it is a reflection of much deeper pandemics in cultural thought about expected gender dynamics. As many in the field of AI prepare for Singularity, we must also develop mechanisms to reflexively address the issues that emerge along the way. Otherwise, the next Women’s March just might be led by Her.
The Invisible Laborer
After panning through its trademark caliginous cityscape, Blade Runner: 2049 introduces K’s love interest, Joi, in its first domestic scene. Though the film is intentionally ambiguous about her personhood at first, it is unapologetically apparent about her role in the house. Following a quaint repartee about cooking dinner and fixing up a shirt, she emerges as a hologram wearing a 1960s updo and a circle skirt that could only have belonged to Donna Reed in a previous incarnation. As the futuristic housewife saunters toward the camera, our suspicion is confirmed: the delectable dish she put on the table is a hologram, too. But we, along with Gosling, continue to play house anyway. Why?
The answer lies in how we conceive of our own labor. When we discuss artificial intelligence, the most common anxiety revolves around jobs. Much of the discourse is predicated upon the premise that some professions will survive the popularization of AI and some will disappear. Certain tasks are valued and others valueless. It is those “valueless” jobs that we want AI to do. At first glance, this seems to make sense. Of course some tasks are less desirable and unworthy of human effort. Why would we not want someone to farm, cook, drive, clean, and free us from these burdens? Once we resolve the job loss in those sectors, the paradigm is bound to evolve into an utopia.
This kind of thinking is not fundamentally flawed, but it is incomplete. It is a natural reaction to grow anxious about the future of one’s job security when the most imminent prospect is automation. However, this angst can also help us reflect on the invisible structural forces shaping our own labor. Frequently, the work that is unvalued is also work done by the impoverished and disenfranchised. Most jobs we are relegating to robots are considered tasks with little to no social value. In turn, people that perform those tasks currently seldom receive recognition or status, socio or economic. In dystopian depictions, there is evermore an enslaved class — underlings who perform requisite tasks that no one else deems worthy. They are embodied by the hooded women in Margaret Atwood’s The Handmaid's Tale, rusty droids in Disney’s Wall-E, and female clones in the year 2144 chapter of David Mitchell’s Cloud Atlas. In addition to performing undesirable labor, these groups face abject discrimination and inequality. Somehow, while we are painting tales of the future with flying cars and holographic companions, we struggle to envision a scenario where work performed by these groups is equally respected.
In a world measured in conspicuous capital flow, those who labor outside of it are rendered invisible. House chores are not work. Grocery shopping is not work. In fact, these biases are so deeply ingrained in our value system that we dare not imagine a future society, accelerated with the aid of AI, functioning in any different way. Of course, this does not mean that automation should be thwarted. It is simply to say that the way that we conceptualize work and the nature of it is fundamentally limited to the status quo. In this framework, it would seem, the importance is that that toilet gets cleaned, and whether the cleaner is a robot or an immigrant is merely a difference in cost. Conveniently, AI allows us to perpetuate this mindset and ignore how societal structures need to change, adapt, and evolve.
The asynchronicity between cultural progress and technological advancement is not unique to AI; similar mismatches have accompanied many prior leaps in automation. Writer and activist Betty Friedan writes in her 1963 book, The Feminine Mystique,
AI cannot just be about efficiency or convenience, and productivity as measured by capital is neither a virtue nor the norm. In most depictions, AI is a not-so-opaque simulacrum that fills the same echelons currently occupied by women, racial minorities, and immigrants. It is a borrowed narrative, stolen from our own realities.
The Despotic Overlord
Despite the two reductive yet pervasive storylines illustrated above, Singularitarians and their cinematic imaginations fear one kind of AI trope the most: the despotic overlord. The narrative seems to believe that technological preoccupations mostly fall in two categories: inventing the intelligent machine itself and contemplating how to avoid our own inevitable downfall. The former decorates magazine covers while the latter haunts our collective psyche. The deep-seated assumptions that a species more intelligent and capable than Homo sapiens will invariably seek power and dominion is overwhelming. Out of these assumptions, doomsday thought experiments like Roko’s Basilisk
In the testosterone-fueled universe of science fiction, fear of AI as a destructive and malicious force runs rampant: from The Terminator to The Matrix and from blockbusters like I, Robot to independent films such as Ex Machina. In these survivalist narratives, AI is developed as assistants to human endeavors and evidence of human ingenuity. As a direct consequence of imposed servitude, the machines’ inevitable malfunction combined with their super intelligence leads to the desire to harm humanity in their pursuit for power and dominance. Yet, must we assume that a hyper intelligent and sentient species will necessarily evolve into despotic overlords? Must the relationship between our progeny and their technology be one of subjugation?
This unease can trace its provenance back to our own assumptions about power structures. Nicky Case
Humans have, through cycles of trial-and-error (and lots of errors, too), largely consented to the virtues of equality, collaboration, and democracy. And yet, even as our societies push towards systems of equity and balance, we still choose to conceive of a comparably intelligent force in the fundamentally limited mode of cut-throat competition where only one winner can thrive. If AI is meant to simulate the better quadrants of humanity, is it not more likely to replicate and ameliorate the success of equal and democratic power structures? Today’s AI is mostly and sometimes solely depicted as in a fiercely survivalist competition with its human counterpart. Even in domains with few pugilistic tendencies, AI is seen by default as an adversary rather than an ally. Case
From historical epics to contemporary headlines, we see the lineage of one dominant theme: us versus them. Believers triumph over heretics. Invaders supplant the indigenous. Democrats against Republicans. New Yorkers versus Bostonians. Based on a few of the darkest episodes in the Anthropocene, it almost appears that the only way humans can make sense of a multitude of value systems is by suppressing all but one. Rather than opening channels of freely exchanged ideals, the current is expected to flow only one way. However, there are also episodes, especially more recent ones, that indicate a trend towards collaboration and mutual augmentation. International alliances, open source technologies, and movements like HeForShe are indications that forces previously thought of as oppositional and territorial can actually blur their own perimeters and become porous and inviting. Sociologist Richard Sennett, in his essay The Open City,
In heeding Ito’s advice about resisting reductionist approaches to Singularity, it is imperative that we recognize our own assumptions about power. Many current narratives focus on myopic self-gains rather than long-term co-prosperity. Artificial intelligence will be smart, but we can choose to imagine that this intelligence will be able to accommodate and learn from multi-axial values rather than having to oppress them. This requires an expansion of our own values and a shift from competitive, win-or-lose paradigms to collaborative, win-win ones. AI derived from a synergetic mindset will most certainly not become despotic overlords. Instead, it will be our partners. Rather than being trapped in the binary of having to either kill us or sweep for us, it would share the workspace, the dinner table, and maybe even the Netflix password.
Most technologists believe that the advancement of AI will result in a better society. I believe it, too — not only in the sense that filing taxes will be easier and chores will be a relic of the past, but also that the process of developing AI will reflect, for our own sake, some of the flawed ways that societies function now. As we sprint to create a new network of intelligence, we ought to first see the problems and imperfections of our own. In fact, current big data endeavors are already revealing structural cracks in our system and painting concrete pictures of previously nebulous biases. Like the Wicked Queen’s smart mirror, scientific advancements should not merely showcase technological capabilities; they must also reflect the assumptions we make and the flaws in the logic.
Frequently, skeptics ask if these technologies will strengthen equality or lead to technocratic extremes. This view assumes that we have to first wait for the technology to mature before we can answer that question. This is not true. The course of developing technological narratives gives us a unique mirror with which to examine our own values. Donella Meadows’ Leverage Points: Places to Intervene in a System
A typical chilling forecast of AI is that it will be smarter, stronger, and more powerful than us, but the real fear should be that it might not be better. It could be instilled with values from our past, with less nuance, more bias, and replete with reductionist tropes. As automation grows, we need to take frequent intermissions to look into the mirror and examine the images it reflects. These technologies are supposed to be harbingers of great scientific progress. Let there be social strides, too.
Bennett, J., & Jones, D. (2018). 45 Stories of Sex and Consent on Campus. The New York Times. The New York Times. Retrieved from https://www.nytimes.com/interactive/2018/05/10/style/sexual-consent-college-campus.html
Case, N. (2018). How To Become A Centaur. Journal of Design and Science. https://doi.org/10.21428/61b2215c
Friedan, B. (2010). The Feminine Mystique. W.W. Norton.
Haas, T., & Westlund, H. (2018). In the post-urban world: emergent transformation of cities and regions in the innovative global economy. Routledge.
Ito, J. (2018). Resisting Reduction: A Manifesto. Journal of Design and Science. https://doi.org/10.21428/8f7503e4
Meadows, D. (1999). Leverage Points: Places to Intervene in a System. The Sustainability Institute.
Palfrey, J. (2018). Line-Drawing Exercises: Autonomy and Automation. Journal of Design and Science. https://doi.org/10.21428/cb377052
Slavin, K. (2018). Design as Participation. Journal of Design and Science. https://doi.org/10.21428/a39a747c
The dawn of artificial intelligence. (2015). The Economist. The Economist Newspaper. Retrieved from https://www.economist.com/leaders/2015/05/09/the-dawn-of-artificial-intelligence
Tolentino, J. (2018). The Rage of the Incels. The New Yorker. The New Yorker. Retrieved from https://www.newyorker.com/culture/cultural-comment/the-rage-of-the-incels