:: Article

reconstruct this: consolation and neural networks

By Amber Husain.

In the winter of 2016, broken hearted in a concrete office block, I was briefly distracted by a ghost. Maria Callas, a column of light in a filmy red dress, shivered in and out of the dark and blasted the corridor with arias from Cherubini’s Medea, assuming, of course, the title role. For two and a half thousand years that mythic witch has reincarnated repeatedly on page, stage and screen to take vengeance on her husband. In the end she usually murders their children.

There was something soothing in the presence of that murderous phantom. At the time I liked to identify primarily as a woman scorned, and tragedy’s familiar script seemed to vindicate the hilarious violence of my misery. But even had my powers of self-dramatization been weaker, there was a level on which any spectre would have done. The hologram (it was a hologram – at first shockingly luminous) flashed an affirmation of technology’s power to crystallise a being – their story, their priorities – in time; to reconstruct to the point of resurrection what was, according to all common sense, lost.

The office block in question was 180, The Strand in London, a repurposed gallery space around the corner from my own, functioning office. Maria Callas – Medea – was herself played by French artist Dominique Gonzalez-Foerster (DGF), who described the piece, Opera (QM.15) (2016), as ‘a kind of séance’. Opera joined a succession of works in which DGF had resurrected fictional or historical figures in spectral form, a project she attributed to her interest in early photography and cinema, and a taste, after her nineteenth-century influences, for the uncanny. The connections made sense. Throughout the mid- to late nineteenth-century and on into the First World War, a cynical industry crept across Europe and America as self-styled ‘spirit photographers’ claimed to have captured on camera the souls of the departed. The spirit portraits they hawked to grieving punters served as visual ‘proof’ of the continued existence of lost loves, summoned via séance or simply ‘there’. Double exposures and composite images conjured a range of ghost forms from milky faces to trails of cloud, in step with the solid living.

If séances are less of a therapeutic staple in the twenty-first century, and the ‘spirit’ element of photography no longer a requirement of its consolatory power, the drive to brandish visual records as proof of something nebulous remains compelling. If, in 2016, I clicked like a metronome through a desktop file of images from the previous half decade, the motivation arose not from any need to prove the continued existence of their subjects’ souls, but certainly from a need to affirm the significances of a life that now felt previous. Such comfort was, obviously, elusive. To trace, in a photograph, the line between my former boyfriend’s face and my own, as though to isolate and circle the point on that vector that signified love, felt as dumb as it was manifestly desperate. A child of the postmodern era, I should have known better than to search for objective meaning in an image.

In 1871 William H. Mumler, the first (and arguably best-known) spirit photographer in the US, was tried for fraud after some of his purported phantoms were recognised still among the living. But if the incident was damaging to a collective belief in ghosts, faith in photographs as fonts of objective knowledge remained, for a time, a dominant position. Implied in the charge of fakery was a belief that photography’s proper purpose (such as Mumler had corrupted) was to reveal and conserve what is ‘real’. In 1931, Walter Benjamin marvelled at how the subject of a photograph is seared by reality with a ‘spark…of the here and now’ – a moment of some truth that, once forgotten, the photo can restore to us. Benjamin compartmentalised the doctoring of photographs in his lifetime, through ‘all the arts of retouching’, as something of an anomaly. But in the daily life of the digital era, we bathe in images not only with the assumption of their having been ‘doctored’, but with a sensibility of awareness that images of any kind are always, in some way, false. Extracted from time and space, the camera’s constructs, from snapshot to durational film, are inevitably, indisputably, constructed.

DGF’s Medea hologram stirred in me a panoply of associative thoughts – a history of signs I could grasp. I knew that Callas had played Medea in Cherubini’s opera through the ’50s and ’60s. I had also seen her play that role in Pasolini’s film of 1969. I knew that both had taken cues from Euripides and, like any good classicist, I knew that play. But no sooner had I been gripped in Medea’s vengeful maw than the Callas ghost morphed into La Gioconda, who was singing something different – something about suicide. The flimsy image fast began to show its seams, and the ‘uncanny’ was back from the land of ghosts. In contemporary discourse, the much-cited ‘uncanny valley’ – the dip in correlation between an artificial being’s human-likeness and our emotional response to it – is observed at the point at which that being is found to be peculiarly visually convincing, but not quite convincing enough. A warning sign might be some mismatch between appearance and behaviour, like expressing surprise with the mouth but not the eyes. DGF’s Callas was unnerving in more than just the facial disjunct between artist and singer. There was also the gap between the shadowiness of the phantom’s body and the clarity of her voice; between the size of the concrete corridor and the sound-tracked murmurs of a vast auditorium; between the youthfulness of Callas’ tone and the distinctive costume of a very late performance. Every glitch sounded some unhappy caution of the recording device’s knack for betrayal.

In the context of solace-seeking, it is no surprise that visions of the uncanny are unsettling. The encounter with an appearance that is somehow subtly ‘off’ threatens the trust with which we clutch at visual records in moments of yearning for certainty, be it about the dead, the distant, or the no-longer-calling. Perhaps such a lack of scepticism is justly punished with unease, the stimulation of which has been Dominique Gonzalez-Foerster’s project since the 1990s. She, together with fellow French artists Philippe Parreno and Pierre Huyghe, have repeatedly worked with contemporary media forms to expose the hypodermic machinations of ideology and narrative in their formation. Eighteen years after its conception, I recently encountered a re-mounting of their animation project No Ghost Just a Shell (2000) at Tate Modern. Technologies of animation may have advanced since the year 2000, but the premises of the project remain relevant. The work revolved around a Manga character named AnnLee. Huyghe and Parreno had purchased her from a catalogue of off-the-peg figures for use in comics, films, video games and adverts. AnnLee was cheap, the price of a character being calibrated in proportion to its complexity. Intentionally left blank for the buyer’s development, she was the perfect prepubescent cipher for Huyghe and Parreno to rent out to other artists, including DGF, for ‘filling in’.

In Parreno’s 4-minute video for the project, devised in an animation facility available to all participating artists, AnnLee comes to life to inform us of her status as ‘a product, freed from the marketplace I was supposed to fill’. ‘I belong’, she says, her voice suddenly drastically altered, ‘to whoever is able to fill me’. DGF’s contribution contains not only two voices, but two AnnLees – Japanese and English-speaking iterations fading into each other while incanting tones of apocalypse. ‘There will be no safety… as sure as my name is AnnLee’. That her name is the only surety available as we trace the girl’s manipulation from one screen to the next testifies to her status as a ‘shell’. Her appearance bears no marks of personal history or symbolic meaning. Huyghe’s version of the character articulates herself as a ‘deviant sign’ – a signifier with no significance, no soul that might qualify her lingering presence for the status of ghost.

No Ghost Just a Shell was initiated in a moment of some culmination for the embrace, in art, of semiotic instability, declared by Derrida and lamented by Jameson as a defining feature of late-twentieth-century culture. By kicking a single image about between commercial and other representational regimes, No Ghost exposed the porousness of image to agenda, and warned against the folly of assuming fixed meaning. Since the millennium, as the bleeding of images into everyday reality has advanced, the warning becomes more urgent and yet harder than ever to heed. Joanna Walsh’s Break.up (2018), a ‘novel in essays’ on the aftermath of a predominantly virtual relationship, describes the unhappy compulsion to live among symbolic relics when all other connections have failed: ‘I have spent too long alone with you – or with your shells: your messages, your photographs, your music’.

Accompanying the decline in my relationship, was a plunge in the degree to which it was conducted in Real Life, and I found in its virtual reconstitution during that period what was, in retrospect, an ideal playground for delusion. Reading into each increasingly rare communication whatever it was that I required, I bypassed any meaning in these that might have prepared me for what followed. Still, when the unhappy truth came, it was again to just these kinds of unreliable missives that I turned for consolation. Walsh’s book gestures in its very existence at the attempt to validate a connection between people who ‘were together in Real Life for hardly more days than a working week’ – a relationship without substance by the standard external metrics. Despite years of ‘real life’ connection, I too felt saddled with a need to reconstruct that bond in a format more visible than memory. Yet, when I tried, I did so with a limp awareness that such ephemera as recollection and emotion were all that meaningfully constituted those talismans I amassed – photographs, videos, screenshots of texts transmitting sentiment no longer felt. The exercise was as empty as the thought of it now brings me shame. All that might have brought satisfaction was the impossible: to have captured and preserved another person’s ‘true’ thoughts, unsoiled by memory, unmediated by language, stored in a box to revisit and wield at my leisure.


However ludicrous my needs might have been in 2016, technology may since have made some incidental leaps in attempting to meet them. Some time since Huyghe and Parreno’s No Ghost mined the Japanese Manga industry as a laboratory of superficial images, an informatics lab in Tokyo has established a practice known as ‘deep image reconstruction’. Its aim, achieved by feeding data from fMRI brain scans into a neural network, is to reconstruct on screen images harvested directly from the head. My exposure to this punchy claim to a feat of mind-reading took place in the perhaps surprising forum of the Serpentine Gallery, via an installation mounted in 2018 by Pierre Huyghe.

The image reconstruction machine was born as follows: people in fMRI scanners were asked to look at various images, and the brain activities corresponding to those images were captured as data. On the basis of this data, an AI was programmed to be able to take in information from new fMRI scans tracking the brain activity of a new person in contemplation of new images, and to generate from these a picture of their mind’s eye. At the Serpentine, Huyghe exhibited on five large screens this process of machine learning – the juddering flicker through possible iterations of an imagination’s contents. What the audience sees is the robotic attempt to land on a final articulation of a man’s visions, bypassing the need for anything so unscientific as his own attempt to describe them.

In 2019, my recollection of the Callas hologram is built from a scaffold of associations. The hand-me-down symbols of Euripides, Pasolini, Cherubini and Gonzalez-Foerster combine as an amalgam in the French tradition of the exquisite corpse, unstable products of human minds offering themselves to a human mind for parsing. The machine responsible for deep image reconstruction tackles interpretation differently, relying on nothing so shaky as association. It plumbs not the depths of a mind’s history, but a set of links, established in the data, between neural activity and structures of colour, line and mass. That such one-to-one links might exist with sufficient reliability to reproduce the image held in any person’s mind is the AI’s most striking proposition, borne out by its relative success in proposing picture reconstructions discernibly related to their original referents. One person’s data goes in so that another person’s vision may come out. Imagination, we are led to believe, might be generic, a spade perhaps officially a spade.

And yet, the question of whether one of the reference photographs reconstructed in this exhibition was, in fact, a spade (we are not shown the originals) is hardly obvious from the shifting mass of diaphanous blobs (the reconstructions) to which we are invited to bear witness. It does, for sure, seem odd to expect patterns of oxygenated blood flow in the brain to capture the imaginative process in its entirety. But say we left the machine’s technical immaturity aside, and took the urge to read and record the workings of other people’s minds to its perfect conclusion. Suppose you were able accurately to scan the imagination at a single point in time, reconstruct the thing imagined with the deepest of mechanisms, download the reconstruction and store it on the hardest of hard drives. Perhaps, you’d think, this process might yield some communication ‘truer’ than the partial products of photography, writing and art. It is still very difficult to derive such a vision from the AI’s onscreen smudge-fest.

To observe the mechanized ‘brain’ at work is tellingly disorienting. As a screen skips through countless permutations of a picture, I am locked in a simultaneous attempt to identify connections between strange shapes and a nameable ‘thing’, but the shapes are shifting faster than I can hold a thing down with a name. My impression of the machine as working in a manner that I can’t quite match only serves to intensify my consciousness of the cluttered arsenal of human expression for which the programme is trying to stand in. As with my reception of Medea’s ghost, the objects I identify with the pictures on screen are grafts of images personally archived; they surface with memory, with abrupt bodily responses, and with whatever happy or unfortunate thoughts trail with them. It is, of course, on such memories and reflexes that the machine has originally fed, and according to such memories and myths that its results will be interpreted. This means of capturing a thought might be artificial and repeatable, but the thought itself is subject to the contingencies of all that is human – embodied and externalised well beyond the confines of the brain. What I perceive to be floating is, to my mother, a submarine. When she sees a vase, I see a woman’s neck.

The ‘deep’ in deep image reconstruction refers to the machine’s ability to decode multiple layers of colour and visual structure – more than the monochrome, linear shapes detectable by earlier forms of technology. What it does not refer to is the kind of depth that might make this method of ‘expression’ more meaningful than what humans themselves can, if only ever partially, achieve. When tasked with handling an unwanted fate, I had thirsted for external means of swaddling myself in the life I already knew, words and pictures as elixirs of contentment, certification of a relationship status I had lost. Standing atop a tower of false memories and elastic vows I stared, through the filter of obsessive digital habits, at the debris of a life that no human communication could help me understand. But faced with the evidence of a drift towards more straight-up channels of communication, early whispers of a possible direct line into the mind, I am filled not only with doubt, but with relief that this doubt looks justified. As life and its priorities change I find, for all I have been able to preserve, that I am grateful, in the absence of an emotional fact-checker, for all I have been able to forget.


Amber Husain is Associate Editor for Art at Thames & Hudson.

First published in 3:AM Magazine: Thursday, May 9th, 2019.