In my shadowy childhood

At the moment, my experiences and memories are incomparable to the experiences of people who are in mortal danger or worried about the lives of their loved ones. But at the same time, I am the person who for many years could not voice my story, which did not fit into the narrative formed by dominant social norms, practices of preserving collective memory, or content filtering algorithms. As a media artist, I am interested in whether AI, which according to digital memory studies could become the "future of collective memory" [1], can help us preserve "voices that are silenced or spelled out by accident" [2].

‍

Now, cut off from the material evidence of my memory, separated by a border I cannot cross, I reflect on the digital frontier where real news about real suffering people cannot be perceived. I turn to studies of digital memory and the role of artificial intelligence in creating collective memory to ask general questions and find specific answers. Reflecting on archival photographs, which can serve as an invitation to nostalgia for a nonexistent past or as documentation of blatant horror, I realize that they were taken by distant witnesses, separated from us not only by experience but also by time. This means that perhaps nothing similar will happen to us. There is a “we” and a “they,” as Susan Sontag rightly noted in her essay “Regarding the Pain of Others” [3]. Perhaps the AI-generated faces of the “COVID memorial” [4]  are too “not ours,” like photographs of those killed in war. Can anonymous victims, this time without racial, gender, or age biases, touch me? Crossing the digital border and entering someone's news feed, these images still cannot penetrate your/our skin, that is, become something real. Or is it that they penetrate too quickly and brutally? The impersonal faces of history that so often appear in media projects about the past—who are they, ghosts that return, shadows that can be mourned, or a convenient container for placing a narrative dictated by power? Purging history of fantasy, dreams, and fears generates a compelling plausible story.

‍

If, as Refik Anadol says, algorithms are capable of "extracting patterns and connections from data, revealing hidden aspects of history and paving the way for unique artistic narratives" [5], then where do they actually get the subjective stories they take into account? It wasn't ChatGPT, but my fellow artists who told me about projects specifically related to intimate stories told through technology [6]. The memory of artificial intelligence is unique in that the data used in the machine learning process possesses its own temporality and spatiality. The further we go back in time or the more we enter "data-poor" regions of the world, the less data is available to "feed" the AI system. This very simple fact makes temporality and spatiality the first and foremost "bias" of any AI system—questions that have been at the core of memory studies from the beginning [7]. Currently, large language models are learning not only to remember but also to "forget" "private information, toxic or sensitive knowledge, and copyrighted content" [8]. Can it apply to collective entities like governments involved in crimes against humanity? Can individuals evoke the right and erase information about their involvement in atrocities as perpetrators or victims? These cases can be declined under public interest, but the balance between individual and collective good in algorithmic remembrance is hard to keep[9].

‍

To have the AI that "changes the future of collective memory," we may need more personal, subjective testimonies and stories, which have even less chance of appearing in someone's news feed than shocking images of war that lure us into a space of hopelessness: "I haven't done this before/I can't fix this/I'm afraid for myself." How can we build digital memory spaces that do not lead to shame for inaction, for preferring some victims over others—spaces that could also offer a chance to carefully extract the shadows of one's past to mourn? How is such a spatial tissue woven, what threads of stories are interwoven into it, how is the amalgam of memories layered, what is the fate of roots that have grown and been uprooted by someone, what digital operations perform all this, and how can we develop a resilient fragility for their perception? In an AI-constructed event, where physicality and locality are absent, such a practice could be, for example, the processuality of private historical reconstruction, which is still quite privileged, as the project "Synthetic Memories" [10] illustrates. While working, ChatGPT suggested that I use the materiality of the visual, treating materials as metaphors, though I feel that the interface procedures of "mixing, flowing, panning and filling" images are felt by us more physically.

‍

The process of reconstruction and independently filling in blind spots implies memory rather than a priori knowledge.

‍

As I delve deeper, pressing the panning button (a Midjourney feature), I try to find artifacts in "memory spaces" that correspond to my "memories" and are not smoothed out by the averaging algorithm. This cognitive operation of expanding and navigating a space made of the “tissue” of the past, forcibly attached to the present, creates a complex experience. In some places, this tissue tears; in others, it forms obscure objects and spaces that resemble a nightmare from which we cannot escape. If we consider artificial intelligence as something that recognizes patterns, then what does it recognize in this case—a random combination of images it must classify according to someone's values, or do I recognize it? Are these images perceived as abstract, distant, and unreal, much like the real images of war from social media that "have nothing to do with you"? By understanding digital memory as a nonlinear processual spatial structure, and carefully guiding myself through a labyrinth of frightening images that are no longer tied to nostalgic landscapes, I hope, following Flusser, to learn to decode the “technical images” [11].

‍

In her text "From Vision to Worldview: Aesthetic Experience as the Tuning of Political Consciousness," Maria Rakhmaninova writes [12]: "Consciousness creates the first assembly of the world in early childhood, in the pre-linguistic era of perception. Without knowing the names or properties of things, it slowly weaves their images into rough patterns and forms an approximate emotional map, marking what is pleasant and unpleasant, frightening and amusing, dangerous and safe, important and not so much. Thus, the visual—alongside the tactile—becomes the very first stop on the way to immersing in the world. Therefore, the analysis of space aesthetics is crucial for studying the structure of mature consciousness and its inherent forms of cognitive activity. Space negotiates between what is known and what can be seen/recognized." Can human culture and imagination learn once again to see the bigger picture of the world within the particulars in digital memory spaces, to recognize the meanings of the images that make up the dataset of one's life, and ultimately learn to differentiate them to overcome cognitive aphasia, see its ghosts, intervene, and resist the automated processes of naming that otherwise remain hidden in the internal circulatory processes of technology? What are the threads of history that cannot be woven into the tissue of the past and present, or are woven too late? How can we feel the present for which there is no language, let alone a common language or translation? How can we accept the present if there is no past? How can we accept the past, the vague shadows of which are forgotten and not fully buried in oblivion—do they appear when the accounts of the dead on Russian social media urge the living to vote for a political candidate in the upcoming elections?  As I look at the "oddities" and "slips" of the images created by Midjourney, I ask: who am I in the space of those holes that will appear in personal and collective history, holes that imagination cannot fill? Am I a node in the network, an element weaving the tissue of communication, the tissue of the present that synthesizes the machine and human dimensions?

‍

In the essay "War Lore," Aimee Walleston writes [13]: "From this, I would argue that the actual content of technical images (however it may presently dictate and dominate our lives) is becoming almost beside the point – what is more important is what is read into them (this coincides with Flusser’s belief that, “Linear texts were originally invented to describe images, and thus served as a function of images, but the relation of this function proved to be reversible, and images were soon used according to texts.”) ...as Flusser points out: “The fateful sentence love thy neighbor acquires a new meaning because the nearer, the more interesting. Love and hatred become integrated within knowledge as they never were during history; one knows what one loves and hates because it is near, and the less something interests, the less one knows it. This is, of course, far more human than the cold, objective attitude of calculating scientific reason. But it is not humane or humanistic: I am far more interested in the fly that bothers me here and now than I am in the future of eight hundred million Chinese, and I love my dog more than I love the suffering masses.” This push and pull of proximity (or at least perceived proximity) is one of the most manipulable aspects of passive reception of the technical image." The right-wing imagination creates virtual emotional sanctuaries for national communities, from which it is very difficult to emerge in the face of so-called external threats, because these external threats—the void of the base, error, and the other—are so frightening to recognize in one's own reflection. Can grassroots practices of remembrance in digital memory spaces and the transformation of memory in new media spaces in general help us emerge from the sanctuary? I hope to see projects about the Soviet nonexistent future as well as the acquisition of an attentive language for understanding the present. I also hope for dialogue because social and cultural memory, as something constantly being born, can hardly be confined to a single country. Forming networks of memorial solidarity among actors in post-conflict societies, the memory activism can go far beyond the local.

‍

Flusser also wrote: 'The dominant media now are images, not texts. A powerful counter-revolution of images against text is taking place.' However, it should be noted that this counter-revolution involves a completely different type of image that did not exist before. The images that program us are post-alphabetical, not pre-alphabetical, like images of the past." In his words, I see many arguments for the idea that the definition of the proximity of an image is a way of observing our relationship with images. The post-alphabetical image is charged with metadata, stereotypes, anger, entertainment, hatred, and assertion. They are not intuitive, do not show the truth, but depend on how deeply we are immersed in their use.

‍

The idea of incorporating sound into my work arose from researching how children learn to speak from birth to one year of age. This is known as language development and is presented as a multidimensional dynamic process closely related to the organism's entire experience. Non-linguistic factors are an integral part of the communicative act and serve as a foundation for the development and continuous use of language. Before verbal symbols are established, numerous sensorimotor experiences occur, and this process is gradual. There is constant interaction between input and output at all levels. A child's receptive language goes through stages where stimulus events are identified, associated, and labeled. Expressive language is viewed as a process of increasing differentiation of outputs. 

‍

Similar to how Hito Steyerl's essay "In Defense of the Poor Image" [14] describes the democratizing potential of low-resolution images, reclaiming visual culture for the people, working with high-resolution memories may become another blind spot. Here, we can attempt to answer the question: what is the difference between a shadow and a ghost? I believe it is due to different contents of the imagination that the ghost haunts and returns repeatedly, while the shadow openly follows you closely and is unlikely to disappear. A ghost may materialize in all its ugliness only to vanish, whereas one must confront the shadow , discovering the boundaries of light and dark. In the context of non-static and changing memory outside the archive, we can imagine not memory itself but a cognitive operation on memory. But who will have access to this "memory interface"? Just as a child learns to speak and remembers new words, memory is formed, reconstructed, and adapted to new experiences and knowledge. Its fluidity, multiplicity, and nonlinearity reflect Deleuze's emphasis on difference and becoming [15]. Constantly changing images and sounds create a tissue of differences that resists static representation. Are we teaching machines to remember, or are they teaching us, or are we conjoined interwoven?

‍

What is the difference between digital memory and algorithmic memory?  At the very heart of technology there is a memory in which something is stored, which is accessed, modified, etc.  The description of algorithmic memory for me is the operation of “content-aware fill”. Not vertical knowledge arising out of nowhere, but "intelligently fills selected areas of an image based on the surrounding content". We can't be sure of its exact referent.  In the context of bubbles of personalized information, the logic of repeating the same thing over and over does not mean memorisation, because the software and events which you subscribed to must be updated to keep the interest of the viewer.  What is remembered becomes a "memorial", a container designed to "calm" and legitimize the ghosts of the past, to give them a place consistent with the existing power narrative of memory. My memory is a cognitive operation of composing a prompt, examining the answer and composing a corrective prompt, and a priori learnt constructions are not suitable for this purpose. 

‍

I remember , when you told me about your experience, it was raining and the city smelled like wet leaves, like a city we may never return to because it was "then".

‍

How do I explain in a  prompt the “smell of wet leaves”, the taste of Proust's "Madeleine cake"? [9] How do I relate algorithmic memory to the qualia that seems to be an integral part of my memory? Today I was watching the news and thinking about worms and the smell of damp soil and I mixed these pictures and further ran a panoramic/content-aware fill  button until I found the very picture we need to forget.

‍

As the developers of Midjourney say: "This is a totally new frontier of image synthesis. Have fun!" [16]

‍