Seeing in an Age of Deepfake: Reflections From One Century Ago

Image designed by Stanley Yang

This famous statistic is now almost a decade and a half old: every two minutes, Americans take more photographs than in all of 19th Century.^[i] Last year, AI created as many images as photographers have taken in 150 years.^[ii] Whichever way we look, considering global data or lived experience, we are voracious makers and consumers of images.

But in a complex visual environment, we should know that “seeing is believing” is no maxim. We read of deepfake’s looming threat evermore frequently, amongst other news that make us question our visual trust. When visuals today are so vivid, voluminous, and vulnerable to edits, sieving meaning from image becomes the puzzle, not the solution.

To prepare for the distortion generative AI impresses upon our information landscape, we should know that the phenomena of image manipulation — that pictorial “truths” can be edited — is not at all new: before AI we had Instagram filters, and for longer still we had Photoshop. Basic and ubiquitous graphic assets such as stock images, Canva-like design tools, and even clip art remind us that the images we often make, use, and see are rarely ever made from scratch.

Those are developments from the digital age, but perhaps more enduring lessons can be mined from even further before. Through a survey of several moments in art and media history, let us consider image manipulation practices from the turn of the last century, and think through the reflections they bear for our AI-inflected, image-inundated age.

Seeing is Believing?

1. Cubism: one image can hold many perspectives

The aim of image manipulation is not always trickery. In fact, for pioneers of a style of painting called Cubism, it was the opposite. Unmoved by the conventions of Still Life and Realism — preceding norms expecting paintings to depict scenes statically and photorealistically^[iii] — artists like Pablo Picasso and Georges Braque wondered if paintings could instead be, and should be, multi-perspectival. For example, even if one was painting a cup, the cup should not merely rest frozen on a table-top and painted from one angle. Instead, the cup could also be depicted from the top, bottom, inside, outside, isometrically, and so on, at once. That ethos intimates the reality that no subject — and no seeing — can be enchained to one singular perspective.^[iv]

Bowl of Fruit, Violin and Bottle (1914), Pablo Picasso[v]

The oil painting above is Bowl of Fruit, Violin and Bottle by Pablo Picasso. What do you see? Each object is refracted into a jumble of parts, and locating items can feel difficult — but maybe, only if one expects to see objects in whole. If we permit ourselves to sense things in parts, to see without expecting a predetermined image, observation becomes easier and we see more: the profile of a musical instrument, newspapers painted over, tassels at the end of a table runner, ornamental carvings on a furniture, maybe a table or a chair, globules of rain from a raincloud. We do see some items specified in the title — but importantly, we also see the larger world beyond.

Descending a Staircase, №2 (1912), Marcel Duchamp[vi]

Multi-perspectivism applied not just to the representation of space, but also time. Above is another painting: Nude Descending a Staircase, №2 by Marcel Duchamp. A model’s continuous walk is captured as a continual series of dynamic poses, as the subject descended from top left to bottom right. The painting captured an arc — not a frozen slice — of flowing movement, across a corresponding arc of flowing time.^[vii]

Today, older paintings in the style of Still Life remain beautiful, and is a valuable exercise for sharpening observation and drawing acumen. Yet, we should remember that hardly anything is “still” about “life”. In the intervening time that a painting took to emerge, the landscape would have changed, the painter would have moved, and several days would have invariably elapsed. Cubist paintings remind me of how our minds works, darting from one urgency to the next, in a world of so many moving and interlocking parts, amidst porous and tangled events in real-time unfurling. An eye hungry only for photorealism might find Cubist paintings distorted, but it was precisely within the folds of that distortion that Cubists yearned to capture reality and perception’s ever-changing heart.^[viii]

2. Pictorialism: photography as the making of, not merely taking of, a scene

Perhaps we less expect paintings to be unalloyed archivers of “truth”, since each painterly stroke embodies human subjectivity. However, film photography is an instantaneous alchemy of physics and chemistry: a camera’s shutter blinks, and in milliseconds light rays surge into the hull of the camera, leaving lasting imprints on a film negative’s silver halide surface. Much of that is similar to how our eye works, through the absorbing of visible light bouncing off the world’s objects. With film photography, it seemed like we had the perfect technology for capturing moments instantly and accurately. The photo felt like a medium with impeccable truth value.

The Raid (1917), Frank Hurley[ix]

And the bar for accuracy is perhaps strictest demanded when capturing the brutalities of war. In 1917, Frank Hurley joined the Australian Imperial Force (AIF) as a wartime photographer, documenting the battles that raged across the Western Front in World War 1^[x]. He produced The Raid that same year. Its content is captivating: troops storming past embankments, charging through a hail of aerial bombardments. Its form is as well magnificent: one is captivated by the photo’s technical and artistic mastery, manifest in its vivid contrast and composition.

It would later be known that the photo was doctored. The Raid is not one photo, but a composite of twelve. Analog technologies might seem tamper-proof, but analog technologies can still be manipulated through analog methods. Hurley was intimately familiar with the techniques of processing image from film negatives, and through precise calibrating of chemicals was able to augment several photos, taken on separate occasions, to look like they were one. Twelve film negatives were selected, spliced, rearranged, and skilfully fused into one image^[xi].

Hurley’s superiors were aghast. Official AIF historian Captain Charles Bean insisted that the images were “fake”^[xii], that they corrupted the integrity of the historical record. But Hurley defended his work fiercely. Capturing war in one shot was impossible. Precisely because war was so wretched and abounding, it was the composite of several brutal images that represented war more truthfully.^[xiii]

Hurley would sometimes be present in public exhibitions of the photo. Overhearing war veterans commending The Raid favourably, he concluded that his techniques were indeed justified.

Stalin Images from 1922 to 1953, reflecting who Josef Stalin considered comrades[xiv]

More flagrant edits have been made, even under the public eye. As Stalin consolidated power during the Great Purge, the removal of those he distrusted can be traced through the multiple versions of his photos. Where there once was an image of Nikolai Antipov, Josef Stalin, Sergei Kirov, and Nikolai Shvernik in a row, each was blotted out by skillful retouchers as they fell out of favour with Stalin, expelled from office or executed.^{[xv][xvi][xvii]}

To edit image is to edit reality — and vice versa. As allies dwindled, only Stalin remained in an increasingly claustrophobic and lonely middle. One imagines that the average person today has tools chief propagandists then could only dream of. Regardless, sophisticated technology is no strict prerequisite for editing; even methods blunt and rudimentary can be powerful.

Photos that involve considerable edits, either when they are snapped or developed, are categorised post-hoc under the term “pictorialism”. Broadly, pictorialism describes the approach — and epiphany — that photography is not a clinical exercise of objective archival, but an involved process of subjective image-making. Just as painters reify paintings with every stroke, photographers imbue emotional, intellectual, and artistic intent into their photos through their own techniques.

Do those techniques make the photos produced “fake”? Or is the real-fake rubric insufficient for unpacking the intricacies of how images are made, and what images can mean?

3. Continuity editing: images and their neighbours story-tell together

Video-capturing technology burst into the scene at the turn of the 20th century. Film editing techniques soon emerged, and we learn that the meaning of an image can be sculpted not only through tweaking what is in it, but also what is around it.

Individually, each shot tells us little. But through the magic of editing, early Hollywood codified an approach for “continuity”: the depiction of believable character living in believable space, with believable intent, in believable time^[xviii]. Consider a simple scene: a young man walks along a riverbank, and something grabs his attention. Editing for continuity, we sequence our shots: first show an establishing shot (of the location, the riverbank), then a mid-shot (of the man’s full body, to show his walking), then a close up (of his face, to show intrigue), and then a reverse shot (of what he was intrigued by, maybe a pamphlet fluttering by, hinting at whatever else beyond).

Still from The Great Train Robbery (1903), directed by Edwin S. Porter. This scene is an example of a “wide shot”, which shows the characters and their surroundings.[xix]

This technique is so ingrained in film grammar that reading the sequence above, one recalls sequences from other films and sitcoms. Individual shots may be simple. But once purposefully stitched, a seed of premise, person, and purpose begins to grow.

Between 1910 and 1920, the summative meaning-making one undergoes when seeing successive images was demonstrated in the classic “Kuleshov effect” by Russian filmmaker Lev Kuleshov. He created an alternating series of images: first the expressionless face of Tsarist matinee idol Ivan Mosjoukine, then something else (a bowl of soup, a girl in a coffin, and a woman on a divan), and lastly the face again. Depending on the combination, audiences perceived Mosjoukine’s expression as hunger, grief or desire.

The Kuleshov Effect (1910–1920) by Lev Kuleshov.[xx]

We feel the effect even when looking at the recreation above, a simplified patchwork made of stills from the short film. We can cognise that Mosjoukine’s expression visually remains the “same” — yet, our minds cannot help but interpret his emotion not as individual image but as a series, influenced by what precedes or follows.

A takeaway here is that image manipulation is not only the act of the image’s creator. It is also us, the image’s witnesses, who re-configure and re-interpret pictures in our minds. Per the Kuleshov effect, even when what we see are unrelated, our minds conjure connections regardless. And today, we almost never see just one image at a time: we see either many at once (on the homepage of websites), or a parade in quick succession (like Instagram stories). In such a visual landscape, what might it then mean to be not just a responsible viewer, but also a responsible and careful interpreter of images?

Seeing is Thinking

We should not underestimate and overlook how disorienting our media and camera-proliferated environment truly are. I think of how our current wars are the first of many to be streamed in real-time: they are broadcast blow-by-blow through body-cams strapped onto combating soldiers,^[xxi] through reporting of and by citizens under siege,^[xxii] through raw footages that seem to indicate authenticity, streamed globally in gruesome high-fidelity…all jostling with cat videos and soufflé. It is obvious why media scholars describe this moment of visual proliferation as an age of “context collapse”.^[xxiii] Multichannel stimuli are dumped onto us devoid of context; even when desired, there is neither footing nor framing to inform our seeing.

But knowing that image-manipulation and visual inundation is not entirely new, that our present moment is neither wholly unprecedented nor ahistorical, can be instructive. We then know that media literacy can and has been learned — consider that early movie audiences, upon seeing an incoming train barreling down on screen, bolted out of the cinema in panic.^[xxiv] Not that that panic was unwarranted; after all, images communicate and convince emotionally, and unlike text are more generous receptacles for our own projections. Pictures reflect the world in our own image, and so imbibing their cinematicism^[xxv] uncritically can entrench biases.

None of that has to be the case. We can develop the instinct of enacting critical pause when interpreting what we see. As AI gets better at seeing and making images, we should also remind ourselves of the depth and complexity that belie image-seeing and image-making.

In that spirit, I find inspiration from title of graphic designer Milton Glaser’s retrospective, “Drawing is Thinking”. No landscape immediately imprints itself onto page; before drawing, the paper is blank. But as one draws, one is actively deciding: what, of the transient inner and outer worlds that I am observing, do I want to commit to the certainty of a drawn stroke? What do I want to memorialise on paper? Drawing is thinking, to be done slowly, and deliberately.

I extend the same thought to seeing, that is to say: “Seeing is Thinking”. When we learn to catch our impulses, we realise that nothing we see has to be imprinted onto our minds unchecked. As one sees, one can actively decide: what, of the transient inner and outer worlds, do I want to pay close attention to? What warrants deeper observation and criticality in my mind?

We can be deeply unsettled by the uncanny realism of Deepfakes today; we can also be sure that tomorrow will upend us again with technologies more bewildering. So more than ever before, our present and future demand our seeing to also be thinking — to be done slowly, and deliberately.

Endnotes

[i] Good, J. (2011, September 15). How Many Photos Have Ever Been Taken? 1000memories blog. Retrieved from http://blog.1000memories.com/94-number-of-photos-ever-taken-digital-and-analog-in-shoebox

[ii] “AI Image Statistics: How Much Content Was Created by Ai.” Everypixel Journal — Your Guide to the Entangled World of AI, August 15, 2023. https://journal.everypixel.com/ai-image-statistics.

[iii] “Cubism was an attack on the perspective that had been used for 500 years. It confused people: they said, ‘Things don’t look like that!” — David Hockney

[iv] “Cubism is like standing at a certain point on a mountain and looking around. If you go higher, things will look different; if you go lower, again they will look different. It is a point of view.” — Jacques Lipchitz
Bork, Bert Van. Jacques Lipchitz: The Artist at Work. New York: Crown Publishers, p. 199, 1966.

[v] Tate. “‘Bowl of Fruit, Violin and Bottle’, Pablo Picasso, 1914.” Tate, January 1, 1970. https://www.tate.org.uk/art/artworks/picasso-bowl-of-fruit-violin-and-bottle-l01895.

[vi] “Nude Descending a Staircase (№2).” Philadelphia Museum of Art. Accessed February 14, 2024. https://philamuseum.org/collection/object/51449.

[vii] “Is it a woman? It is a man? No. To tell you the truth, I have never thought about what it is. What should I think about that? My images are not objects, but rather abstractions. ‘Nude Descending a Staircase’ is an abstraction of the movement. Nobody had thought to introduce a movement, the actual movement, into a work of art. […] The futurists and I had planned it at the same time, it was already in the air somehow because of the invention of film and cinema. The fact that I had seen chronophotographs gave me the idea for the Nude. […] And in the end, a painted picture is a diagram of an idea, and in this case this has never been done.” — Marcel Duchamp

[viii] “Cubism is not a reality you can take in your hand. It’s more like a perfume, in front of you, behind you, to the sides, the scent is everywhere but you don’t quite know where it comes from.” — Pablo Picasso
Podoksik, A. S., and Pablo Picasso. Pablo Picasso: The creative eye (from 1881 to 1914). Bournemouth: Parkstone, 1996.

[ix] Wikimedia Commons. Accessed February 14, 2024. https://commons.wikimedia.org/wiki/File:Frank_Hurley_composite_image_--_the_raid_(over_the_top).jpg.

[x] NSW. Accessed February 14, 2024. https://acms.sl.nsw.gov.au/_transcript/2007/D00007/mss314_p77-96.pdf.

[xi] “Great Australian Photographs: Frank Hurley’s the Raid — an Audio Essay.” The Guardian, May 28, 2017. https://www.theguardian.com/artanddesign/2017/may/29/great-australian-photographs-frank-hurleys-the-raid-an-audio-essay.

[xii] “Truth and Photography.” State Library of NSW, June 30, 2022. https://www.sl.nsw.gov.au/stories/frank-hurleys-world-war-i-photography/truth-and-photography.

[xiii] NSW. Accessed February 14, 2024. https://acms.sl.nsw.gov.au/_transcript/2007/D00007/mss314_p77-96.pdf.

[xiv] Chapple, Amos. “Fake Views: The Good and Bad of Soviet Photoshopping.” RadioFreeEurope/RadioLiberty, July 13, 2018. https://www.rferl.org/a/soviet-airbrushing-the-censors-who-scratched-out-history/29361426.html.

[xv] “How Photos Became a Weapon in Stalin’s Great Purge.” History.com. Accessed February 14, 2024. https://www.history.com/news/josef-stalin-great-purge-photo-retouching.

[xvi] Chapple, Amos. “Fake Views: The Good and Bad of Soviet Photoshopping.” RadioFreeEurope/RadioLiberty, July 13, 2018. https://www.rferl.org/a/soviet-airbrushing-the-censors-who-scratched-out-history/29361426.html.

[xvii] Rhp. “How Stalin’s Propaganda Machine Erased People from Photographs, 1922–1953.” Rare Historical Photos, December 7, 2021. https://rarehistoricalphotos.com/stalin-photo-manipulation-1922-1953/.

[xviii] Bordwell, David. “Intensified Continuity Visual Style in Contemporary American Film.” Film Quarterly 55, no. 3 (2002): 16–28. https://doi.org/10.1525/fq.2002.55.3.16.

[xix] “The Great Train Robbery.” IMDb, December 7, 1903. https://www.imdb.com/title/tt0000439/.

[xx] StudioBinder. “How Spielberg Subverts the Kuleshov Effect.” StudioBinder, June 25, 2021. https://www.studiobinder.com/blog/kuleshov-effect-examples/.

[xxi] Liebermann, Oren. “Exclusive: Bodycam Video Shows Early Moments of Hamas Massacre in Israel and Tunnels under Gaza.” CNN, November 15, 2023. https://edition.cnn.com/2023/11/15/middleeast/bodycam-video-hamas-massacre-tunnels-intl/index.html.

[xxii] Instagram. Accessed February 14, 2024. https://www.instagram.com/wizard_bisan1/.

[xxiii] Brandtzaeg, Petter Bae, and Marika Lüders. “Time Collapse in Social Media: Extending the Context Collapse.” Social Media + Society 4, no. 1 (January 2018): 205630511876334. https://doi.org/10.1177/2056305118763349.

[xxiv] Bottomore, Stephen. “The Panicking Audience?: Early Cinema and the ‘Train Effect.’” Historical Journal of Film, Radio and Television 19, no. 2 (June 1999): 177–216. https://doi.org/10.1080/014396899100271.

[xxv] Garber, Megan. “The Trump AI Deepfakes Had an Unintended Side Effect.” The Atlantic, March 24, 2023. https://www.theatlantic.com/culture/archive/2023/03/fake-trump-arrest-images-ai-generated-deepfakes/673510/.

Jevon Chandra is Foresight Analyst at the Centre for Strategic Futures.

The views expressed in this blog are those of the authors and do not reflect the official position of Centre for Strategic Futures or any agency of the Government of Singapore.