SEPTEMBER 25, 2021
ALTHOUGH 2021 CONTINUES to be a year of unbearable deaths, it’s also a year of uncanny resurrections.
On February 23, the bicentennial of Romantic poet John Keats’s death, I decided to walk to his London home. When I arrived at the long white house in Hampstead, however, there were few signs of the momentous anniversary because of the pandemic. Somewhat dejected, I returned to my flat instead to watch a Zoom celebration that debuted the Institute for Digital Archaeology’s reconstruction of the poet, based on contemporary images and masks, that could gesture and recite his poems, accent and all. Filled with a mixture of wonder and enthusiasm for future improvement of this developing technology (it still bore the geodesic surfaces of many digital likenesses built from scratch), I was mostly warmed by the feeling of collective global appreciation for a poet none of us had met, as well as the potential to revivify him for students and the public.
Then Deep Nostalgia happened.
Just two days after Keats’s bicentennial, the Israeli genealogy site MyHeritage released Deep Nostalgia, a technology developed by Israeli security company D-ID (De-Identification) as part of their “reenactment suite,” that animates human faces in photographs, making them look around, nod, smile, wink, and even dance. Riding my techno-high, I created an account and uploaded paintings of Keats. To my delight, the Romantic poet was soon blowing kisses at me through the screen. It was mesmerizing.
According to MyHeritage, Deep Nostalgia is a way to bring to life photographs of dead ancestors and famous people. The technology took social media by storm, and users have generated nearly 100 million animations, including ones of Frederick Douglass, Queen Victoria, Marie Curie, Mark Twain, Florence Nightingale, Martin Luther King Jr., and Ruth Bader Ginsburg.
I put Deep Nostalgia to the test. It recognizes and animates a wide variety of media, including artworks. Compared to the often-biased standards of current facial-recognition technology, it does relatively well with faces of color, disfigured faces, and faces partially obscured by veils, sunglasses, or the omnipresent coronavirus mask. It doesn’t usually recognize cartoon faces (sorry, Fred Flintstone) nor nonhuman faces (animating your dog will have to wait), and it performs less well on faces in makeup, at unusual angles, or ones with eyes and mouths closed. I tried animating Keats’s life mask, for example, but Deep Nostalgia renders his eyelids in REM. It was enough to make me lose my religion.
As thrilling as Deep Nostalgia is to many users, just as many have been disturbed by it. MyHeritage itself cautions that “[s]ome people love the Deep Nostalgia™ feature and consider it magical, while others find it creepy and dislike it,” calling it “uncanny” — a reference both to Freud’s unheimlich and Masahiro Mori’s bukimi no tani genshō (不気味の谷現象), usually translated as Uncanny Valley.
Deep Nostalgia — along with the bicentennial Keats, Ai-Da, and Woebot — has been just one of the examples I’ve encountered this year of our ongoing efforts to technologically reproduce the human. It’s also one component of the ongoing aesthetics of nostalgia. In music alone, for example, Dua Lipa continues to reissue Future Nostalgia, Jimmy Jam and Terry Lewis are embracing “newstalgia,” and Lil Nas X recently called nostalgia a “lying ass bitch” on Twitter. But these developments — and our reactions to them — aren’t unprecedented. Similar stuff has been happening for centuries.
So Deep Nostalgia made me wonder: Why do we continue to pursue such efforts when the results inevitably creep us out? Why do we constantly retread the past, (re)animating dead loved ones, even when it can be painful?
Basically, why do we keep doing this to ourselves, and is this really about nostalgia?
A Brief Reproduction of Nostalgia’s History
Nostalgia was a disease. At least, that’s how Europeans viewed it for more than two centuries. Antiquity long recognized the anguish of homesickness — think Odysseus at sea or the Jews in Babylon — but modernity pathologized it. Swiss physician Johannes Hofer coined nostalgia (nostos = homecoming, algia = pain) in the late 17th century to describe the enervation of mercenaries who longed for their Alpine homeland while fighting abroad. From the beginning, nostalgia was a spatial disorder: when away from their native habitats, bodies declined. Once pathologized, the disease suddenly seemed to be everywhere. As Jonathan Schroeder notes, the American Civil War was the apex of the pandemic; more people were diagnosed with nostalgia during that conflict than with any other disease.
It’s no accident that nostalgia flared alongside the stirrings of the Industrial Revolution. The proliferation of technologies both caused and was caused by nostalgia. On the one hand, mechanization displaced manual laborers, forcing many of them to leave their farms and villages for urban factories. On the other hand, inventions from the train and telegraph to the telephone and even the player piano aimed to keep beloved people, places, and things within reach. Most prominently, the photograph and phonograph (and eventually cinema) were devised partly to salve nostalgia. By the late 19th century, you could carry a photo of your spouse to the battlefront or hear your parents’ voices after they died. This was the age of the souvenir, keepsake, replica, and memento. The tension between alienating and simultaneously assuaging nostalgia was an essential ingredient in the rise of technology.
Nostalgia seemed to be found most among the displaced because it was conceived as a disorder of displacement. Huge numbers of diasporas — including those of ousted Indigenous populations, enslaved Africans, and persecuted Europeans — afflicted people with longing for their homelands. As so often happens in pathology, prognosis blurred with diagnosis as these groups became increasingly stigmatized. Now, the huddled masses yearning to breathe free were not just afflicted by nostalgia, they came to be seen as biologically prone to it. Whole groups were displaced not just because the powerful subjugated them but because they were deemed fundamentally subjugatable.
By the time photography was widespread in the late 19th century, nostalgia comprised a power dynamic capable of elevating certain groups while dehumanizing others. Not surprisingly, this dynamic grafted onto technology, part of what Svetlana Boym has called “technonostalgia” or the “nostalgia industry.” While reproduction technologies were constantly celebrated in the press, there was also widespread ambivalence. For example, some hailed the phonograph while others hesitated. When a famous recording of Robert Browning’s voice, one of the earliest in history, was replayed in 1890, one year after his death, his sister denounced it as an “indecent séance”: “Poor Robert’s dead voice to be made interesting amusement. God forgive them all. I find it difficult.” (Indeed, the developers of Deep Nostalgia expressly excluded speech to prevent “abuse” of their technology, though speaking animations of Browning and others have existed on YouTube for almost a decade.) In similar fashion, Emily Dickinson described photography as “Death’s surprise, / Stamped visible -,” avoiding cameras because “the Quick wore off those things, in a few days, and forestall the dishonor.” To many in the 19th century, reproduction wasn’t just creepy, it was somehow … wrong.
How can we articulate their aversion — and ours?
The In-Control Valley
We talk about the Uncanny Valley, which we treat as an evolutionary reaction, part of our lizard brains, and scholars are studying this. But guided by our forebears, I think something else is also happening with our feelings about reproduction technologies, which Deep Nostalgia evokes with unusual force.
On the Uncanny Valley graph, we usually plot objects like robots, zombies, and puppets, whereas Deep Nostalgia is designed to animate people who were living when the image was taken. The former focuses on thing as human, the latter on human as thing. This is where nostalgia’s latent power dynamic, which I described earlier, comes into play.
Welcome to the In-control Valley.
I didn’t put it quite this way before, but technology is also about control. At the touch of a button — and it’s almost always a button, that icon of omnipotence — you can brew tea, make a motor car, or bomb a city, as one Cold War–era pop song put it. And yes, at the touch of a button, you can replay a voice long after its owner is dead; capture the image of a relative from another hemisphere; animate a photograph that has never moved before. Imagery in particular attracts a language of control: you take someone’s picture; you shoot a video of someone.
When it comes to nostalgia, one thrill of technology is the thrill of having our power restored. Deep Nostalgia empowers us to conquer time and space to get our dead loved ones back, like Orpheus. But we must remember technology’s double-edged sword: to gain control is to align ourselves with the powerful. Our joy in using Deep Nostalgia is “the joy we feel when fighting alongside the mighty,” as S.P. recently put it. Though everyone bears individualized historical scars, we all nevertheless have in common that we’ve been forced out of our own pasts, displaced in time. We all therefore bear wounds of displacement, even as we use this technology to displace others.
Enter feelings of guilt. Guilt that — to our raw perceptions, anyway — we are controlling dead relatives and famous people. Guilt that we are basically hijacking their free will. And guilt that we are reveling in it. This is where the two valleys differ. The Uncanny creeps us out biologically, but the In-control creeps us out ethically. And the intensity of both maximizes at the nadir of the graph.
Lately, I’ve been preoccupied principally with two effects of Deep Nostalgia’s co-option of free will. One seems to me a sort of rights violation: Deep Nostalgia doesn’t just make people move; it makes them move in ways they never actually did, which is something that earlier iterations of photography, phonography, and even film couldn’t accomplish. Lots of press coverage has compared Deep Nostalgia’s animations to the moving portraits in the Harry Potter series, but Deep Nostalgia is more like the Imperius Curse, which is “Unforgivable” because it overrides free will. This instinctive distaste sometimes surfaces in media news, as when Marvel recently decided not to use CGI to include the late Chadwick Boseman in future Black Panther films, or when fans reacted negatively to an AI reconstruction of Anthony Bourdain’s voice in Roadrunner.
Bound up with Deep Nostalgia’s imposition of movement are complex matters of representation — not just unique facial mannerisms or impairments, but ethical signaling. Some commentators, for example, have pointed out how animating photos of Frederick Douglass is a sticky wicket. On the one hand, the abolitionist was known to pose decorously to avoid the “happy slave” trope, so that making him smile, wink, and dance seems wrong. On the other hand, he probably did do all those things (willingly) while alive; it might have been the strictures of 19th-century society and photography that mainly motivated his choice. Deep Nostalgia, in other words, could enliven historic figures like Douglass beyond these obsolete constraints. As a teacher of 19th-century culture, I can attest that such media deployed carefully in the classroom can help students see the period afresh.
Shoshana Zuboff, a scholar of “surveillance capitalism,” has argued that we need democratic reform to protect our “elemental epistemic rights” — that is, our choice to be known or unknown to other people. These rights are not properly protected, she argues, because “they had never come under systematic threat, any more than we have laws to protect our rights to stand up or sit down or yawn.” But it seems to me those are just the sorts of rights Deep Nostalgia raises: Do we have the right, especially after death, not to stand up or sit down or yawn? In addition to formulating elemental epistemic rights, do we need a necroethics? These questions about consensual animation are one effect of the In-control Valley.
My second issue is more ontological and involves the predictability of Deep Nostalgia’s animations. We tend to think that knowing people, truly loving them, is to be able to predict their every move — you know, the way they wear their hat, sing off key, hold their knife, and all that. But unpredictability might be of greater importance than we usually acknowledge. Being surprised by a spouse’s reply or bumping into an old friend or, yes, being shocked by a relative’s death manifests their realness to us, confirming that this other person is autonomous because we never could’ve guessed how they’d act. This is, so to speak, how we humans pass the Turing Test. In short, reality is the thwarting of our expectations.
To me, Deep Nostalgia does away with that spontaneity by the stroke of an algorithm. Everyone from Amelia Earhart to Aunt Ethel does exactly what we expect because we choose exactly what they’ll do. This effect of the In-control Valley is quite like that of pathologized nostalgia in which diagnosis became prognosis: if predictability makes someone seem less real, then the people in these animated photographs become less real because they’re predictable.
So what are we to do … if anything?
The Uncanny Asymptote
Reproduction redefines. When we make copies of something (an image, sound, movement), we permanently alter it and its relations to the world. Walter Benjamin’s name for this phenomenon was aura. Frequently, aura is understood to be the native environment of an artwork. Think The Creation of Adam in Michelangelo’s Sistine Chapel fresco: if you reproduce that image on a T-shirt or a coffee mug, you remove it from its original environment and intention and thus destroy its aura. (In fact, loss of aura is what makes a souvenir a souvenir, or as Susan Stewart puts it, “The souvenir by definition is always incomplete.” But that’s a long-ing tale for another time.)
There’s another aspect to aura, though: a copy creates the original. An object only becomes an “original” once copies have been made of it. People pay good money and endure airplane meals to see the Sistine Chapel in person precisely because they have seen The Creation of Adam reproduced countless times in books, on screens, and yes, on shabby souvenirs. That’s one way an image manifests its aura.
At first, I thought that aura was bad news for Deep Nostalgia. Digitizing, enhancing, colorizing, and animating a photograph strips away its aura. But I’ve come to think that aura’s other aspect — that a copy creates the original — might help us see something a little surprising about what it means to be human. Here, another foray into the past suggests that we possess this curious instinct to move the goalposts of what it means to be human each time a technology like Deep Nostalgia comes along.
Jessica Riskin, writing about Enlightenment Europe’s craze for automata (robot-like machines designed to reproduce human biology), observes that the
contradictory convictions — that one could understand life and intelligence by reproducing them, on the one hand, and that life and intelligence were defined precisely by the impossibility of reproducing them, on the other — went into operation in the early part of the eighteenth century. […] The result was a continual redrawing of the boundary between human and machine and redefinition of the essence of life and intelligence.
This continual redrawing of boundaries coincided with the rise of technologies that tango with nostalgia.
If we take Riskin’s observation and cross-apply it to reproduction, we find that as reproductions become more lifelike, we redraw the boundaries between a human and what constitutes a “successful” replication of one. For example, once upon a time, Renaissance paintings strove to reproduce a face’s contours, its light and shadow. Photography (literally “drawing with light”) was invented to do exactly this. Once photography became proficient at capturing contours, the boundaries between a face and its reproduction shifted. Now photography’s grayscale became an obvious deficiency, and color was a compelling reason to see someone in person. When colorization was achieved, the boundaries shifted again. Deep Nostalgia’s animations will shift the boundaries once more. (By contrast, a recent art project reverses this process, applying general adversarial networks to family photos to produce new painting-like images.)
I wonder if this technological history is why so many Deep Nostalgia animations in news coverage have been of 19th-century figures. Most of that century was a Goldilocks zone in which photography was common, but animation (e.g., cinema) wasn’t. We have copious period images but few movements. Deep Nostalgia is at its most impressive in this zone because we’ve never seen Abraham Lincoln or Charles Darwin move. Since there’s nothing to compare these animations to, we deem them miraculous. By contrast, when we see MLK or RBG animated, we’re prone to mock the effect since we’ve all seen their facial movements, and Deep Nostalgia falls short.
Perhaps we should amend the Uncanny Valley graph then. Since we keep nudging the righthand edge — which represents us, the certifiably human — the curving line is asymptotic. As it ascends, it’ll come closer and closer but, by definition, never reach us. We instinctively make sure of that.
So move over, In-control Valley. Make way for the Uncanny Asymptote.
If the Uncanny Asymptote exists, Deep Nostalgia could prove a surprisingly humane philosophical aide. Our instinct to keep the gap pried open as if it were an escape hatch might seem like a pathetic rationalization for preserving human exceptionalism; maybe when push comes to shove, we really will give up on ourselves in the future because reproduction is maximally successful, and that curving line will pierce the graph’s edge. After all, there’s been much fretting recently about deepfakes, even historical ones — Deep Nostalgia’s siblings — becoming lifelike enough to dupe or replace us. But I think something else might be going on.
Perhaps all this time, the long line of reproduction technologies — of which Deep Nostalgia is a scion — has effectively helped us hone “the human,” paring extraneous requirements away from meaningful existence and interaction. Partly due to this zeroing-in on essential humanity (alongside our constant political reforms), being human in many societies at least purportedly no longer requires being male or white or land-owning or straight or cis or able-bodied or sane or innocent or …
Or we still have a long way to go. Obviously.
There’s no denying technology has had a huge hand in subjugation and dehumanization. But I think the In-control Valley and Uncanny Asymptote might also help us toward our destination, the former because it viscerally forces us to acknowledge the power dynamic inherent in technological reproduction, and the latter because it instinctively forces us to recognize a more basic humanity.
For me, Deep Nostalgia debuted at a propitious time. As we stagger out of our own 21st-century pandemic, full of longing for prior worlds that did and didn’t exist, now is a good time to redraw boundaries and redefine ourselves again. My own revised definition might be something like: Humanity consists in that which we aspire to reproduce but cannot.
At least, that may be “all / Ye know on earth, and all ye need to know.”
Justin Tackett is a Leverhulme Early Career Fellow in the Department of English and Comparative Literary Studies at the University of Warwick in the United Kingdom specializing in literature and technology of the long 19th century. His website is www.justintackett.org and his Twitter handle is @justinctackett.