Calling a camera blind appears to be a paradox, given how we conceive of cameras as devices that capture light and document them as images. Cameras—and the practice of photography—have always been closely associated with the phenomenon of sight.

Derived from the Latin “camera”, meaning a chamber or small room, the word has likely come to mean a photographic device today through its precursor, the camera obscura, which refers to a dark room where a small hole at one side is used to project an image of the view outside onto a wall or surface opposite the hole. That said, this etymology of the word “camera” suggests a more expansive scope, for the space encompassed by a room need not solely be seen, but can also be felt or heard.

In this regard, Diego Trujillo Pisanty’s Blind Camera (2022) is aptly named, for it pushes us to expand our conception of what a camera can be. This project centres on a unique AI-powered device created by Diego that plays on the function of a point-and-shoot camera. Instead of creating pictures from light, Blind Camera generates images from sound at the push of its button. At the heart of this process is an artificial neural network custom-made and trained by the artist to find a common representation between a sound and an image. With videos taken in Mexico City, where Diego is based, the network was trained on the association of each video frame with its previous second of sound. As the artist described on his website, the network would “(1) encode the sound into a vector, (2) decode it back to the matching image, and (3) try and convince another network that the resulting image is a photograph.”

The result of this reimagination of the camera is a series of 100 images, which was released and sold as part of Fellowship’s “Post Photographic Perspectives III: Taming The Machine” exhibition in April 2024. Each image in this series seeks to give visual expression to the cacophony of sounds that engulf Mexico City today, transfixing the musicality of its sonic landscapes into the stillness of a static image. In parallel, the 100 individual works that make up Blind Camera were also minted as NFTs on the Ethereum blockchain, thus preserving certain aspects of these works, e.g. their transaction-related data, within the recesses of a decentralised and thus more durable database.

What drew me to collect two works from Blind Camera was its ingenious concept. I found Diego’s use of sound as the basis of image generation immediately compelling and refreshing, especially when seen in the context of the recent emergence of text-to-image generative AI tools.

Such text-to-image tools are undoubtedly becoming incorporated as part of legitimate artistic practices focused on the relationship between the verbal and the visual—a relationship long-appreciated by the generations of storytellers, poets and writers who have conjured vivid, intricate imagery through the words they speak or write. Nevertheless, as much as text-to-image AI tools are poised to redefine humanity’s visual languages, Blind Camera reminds us that they do not represent the only way that AI can be harnessed to create visual art.

https://opensea.io/assets/ethereum/0xFe4e82E80D1e70c6B60eDE19A8BF0B63F5c1b4F3/122

Words, while adept at conveying multitudes, have never held a monopoly on defining our imagination or gaze. We have always relied on the full, multi-dimensionality of our sensory toolkit in our creative processes—synthesising the varied sights, textures, aromas and melodies around us into things that we want to see and to be seen. For example, Synchromism was a movement in painting in the early 20th century based on the idea that colours in a painting could be arranged in a similar manner as notes in a musical piece. Indeed, sound and other sensory inputs have rich visual associations, and Blind Camera can be appreciated as part of a longer tradition of artists tapping on these associations to create works of art.

In using AI, Blind Camera certainly stands out in terms of its methodology to express the relationship between sound and imagery. That said, it is not the use of AI per se that gives this work its conceptual heft.

What made me pause when looking at the images created by Blind Camera is realising that the series represented an entire system of perception—a novel way of seeing—from what we’re typically accustomed to. In Diego’s own words, each image is “indexical of the surrounding soundscape and not of the scene in front of the camera.” But I also think there is more to this: each image in the series was not the result of an individual’s gaze (or hearing) at a specific moment, the basis of a point-and-shoot camera, but an agglomeration of different sensory inputs that span across time and space. The works are pictures, yet more than pictures—not only because they were created from the soundscape of a particular place, but also because their expression was mediated through a synthesis of multiple sounds from multiple places, i.e. the training of its underlying AI model.

Chromesthesia may provide a useful frame of reference here. Also known as sound-to-colour synaesthesia, individuals with this condition can perceive colours, shapes and movement involuntarily when hearing sounds. Such perceptions, however, would be idiosyncratic—different individuals with the condition would likely experience the same sound differently. As someone without chromesthesia, I would also never be able to experience how someone with the condition could perceive sound, beyond what their own descriptions or expressions conveyed. The system of perception as represented by Blind Camera may thus be the best possible means for me to experience chromesthesia directly, and as a more objective phenomenon that goes beyond a particular individual’s experiences. In this sense then, the images of Blind Camera may reflect a collective form of chromesthesia—the aggregate of the myriad ways in which sound and imagery can be associated together.

https://opensea.io/assets/ethereum/0xFe4e82E80D1e70c6B60eDE19A8BF0B63F5c1b4F3/125

Beyond the conceptual novelty of Blind Camera, what has sustained my interest in the work over time is how it grounds me intimately in the local—the particular rhythms, textures and atmospheres that Mexico City encompasses, an “enormous improvised hypermetropolis” (per David Lida’s description of the city in his book, First Stop in the New World) heaving with almost ten million inhabitants and many more on its periphery.

The blurry, undeveloped yet evocative elements in the images spanning this series invite one to speculate on how the actual scene behind each piece would have looked like. One can almost imagine masses of people, vehicles, buildings within the images. Some seem to drown within the din of Mexico City, while others seem to find a space to emerge with clarity. Indeed, the limitations of this “blind” camera beckon and instigate us to look closer—to cast our own gaze towards the same subjects and try to illuminate them in our mind’s eye.

Naturally, I felt this impulse most strongly when I was physically in Mexico City, travelling there with my wife for a week last month. As we walked around different neighbourhoods within the city and took in its many sights (and sounds!), I couldn’t help but think about where Diego could have captured the sounds used to create the 100 images in the series—would my physical proximity to their underlying soundscapes allow me to see through the images?

Nevertheless, I soon realised that this train of thought was futile, for each image was not a reflection of the sensory landscape of the city but a refraction. Even though each of the 100 images was created based on the soundscape of a particular place at a particular moment in time, each of them should be more accurately seen as a composite built upon a larger set of sonic and visual data drawn from the city. The point of Blind Camera was not to recreate the sensory reality of specific localities across Mexico City, but to remix them through the specific medium of AI, taking in both its unique affordances and constraints.

In this regard, the artist’s hand in Blind Camera is most salient not in the decision on where to take the “snapshot” for each piece, but in what other data he has collected to train his AI model. After all, even though each image was seeded by the specific soundscape of a particular site, this sensory input would have been mediated by the “learnings” of the AI model gleaned from other sensory data from other sites within the city. That we can create an entire system of perception based on this process in the first place reveals the tremendous power of AI to find patterns and synthesise data. Yet, this also demonstrates the subjectivity of AI, for what was used to train the models ultimately defines the boundaries of their own “dark rooms”—the latent space from which they generate the things we prompt them to.

This subjectivity, however, should be seen as a feature and not a flaw for art created with AI. In the case of Blind Camera, we should not rely on it for an “objective” reimagination of Mexico City, for an individual’s experience of cities has tended to be highly personal. Art’s role in expressing or critiquing the character of a city is thus different from academia’s—not to tease out some objective truths from the writhing, burgeoning urban environments that increasingly define the contemporary human condition, but to marinate within its subjectivities and make us reflect on them in our own way. The use of AI simply injects another layer of subjectivity into this process.

That the Mexico City depicted through Blind Camera is only a subset of the city that the artist has seen and heard is thus to be expected. What is more crucial is whether the artist has left sufficient space for their works to transcend the subjectivity of their individual positions within the city.

On this, Blind Camera shares some similarities with Francis Alÿs’ Colector (1990-1992), in which the artist walked through the streets of Mexico City, pulling a small magnetic toy dog that collected various metallic debris along the way. Just as how the material output in Colector is dependent on Francis Alÿs’ walking path, the latent space of Blind Camera’s AI model is likewise bounded by the data of the city that Diego has collected. That said, in both works, we cannot see the full extent that the two artists have traversed the city spatially, beyond blurry snippets of where they might have been.

To fully appreciate the scope of their works then—the tangible products of their walks—we thus have to imagine the artists’ physical journeys ourselves based on our own impression of the city, whether we have been there physically before or not. What we make out of the detritus picked up by Colector or the images generated by Blind Camera will thus be shaped by our personal preconceptions and experiences of Mexico City as much as the artists’ own.

Indeed, Blind Camera provides a prism through which one can perceive Mexico City, where the local is refracted with the personal—the subjectivity of the artist melding into one’s own. In a way, it is a camera turned onto itself. What we see in the images it creates depends on all the prior sensory data that has gone into making this camera. This in turn prompts us to speculate on the artist’s journey through Mexico City, and inevitably leads us to consider our own relationship with the city.

For those who have never been to Mexico City, it will be worthwhile to keep Blind Camera in mind when you do have the opportunity to visit. The city can be incredibly overwhelming, inundated with multiple layers of history, complexity and nuance. In this respect, a work like Blind Camera, grounded in the intricacies of the city, will provide a handy accompaniment as you try to make sense of what you see and hear in this place that the Aztecs once conceived as the navel of the world.

https://opensea.io/assets/ethereum/0xFe4e82E80D1e70c6B60eDE19A8BF0B63F5c1b4F3/119

Credits: The header image of this essay was cropped from Blind Camera #4, one of the two pieces from this series that I have collected.

Calling a camera blind appears to be a paradox, given how we conceive of cameras as devices that capture light and document them as images. Cameras—and the practice of photography—have always been closely associated with the phenomenon of sight.

Derived from the Latin “camera”, meaning a chamber or small room, the word has likely come to mean a photographic device today through its precursor, the camera obscura, which refers to a dark room where a small hole at one side is used to project an image of the view outside onto a wall or surface opposite the hole. That said, this etymology of the word “camera” suggests a more expansive scope, for the space encompassed by a room need not solely be seen, but can also be felt or heard.

In this regard, Diego Trujillo Pisanty’s Blind Camera (2022) is aptly named, for it pushes us to expand our conception of what a camera can be. This project centres on a unique AI-powered device created by Diego that plays on the function of a point-and-shoot camera. Instead of creating pictures from light, Blind Camera generates images from sound at the push of its button. At the heart of this process is an artificial neural network custom-made and trained by the artist to find a common representation between a sound and an image. With videos taken in Mexico City, where Diego is based, the network was trained on the association of each video frame with its previous second of sound. As the artist described on his website, the network would “(1) encode the sound into a vector, (2) decode it back to the matching image, and (3) try and convince another network that the resulting image is a photograph.”

The result of this reimagination of the camera is a series of 100 images, which was released and sold as part of Fellowship’s “Post Photographic Perspectives III: Taming The Machine” exhibition in April 2024. Each image in this series seeks to give visual expression to the cacophony of sounds that engulf Mexico City today, transfixing the musicality of its sonic landscapes into the stillness of a static image. In parallel, the 100 individual works that make up Blind Camera were also minted as NFTs on the Ethereum blockchain, thus preserving certain aspects of these works, e.g. their transaction-related data, within the recesses of a decentralised and thus more durable database.

What drew me to collect two works from Blind Camera was its ingenious concept. I found Diego’s use of sound as the basis of image generation immediately compelling and refreshing, especially when seen in the context of the recent emergence of text-to-image generative AI tools.

Such text-to-image tools are undoubtedly becoming incorporated as part of legitimate artistic practices focused on the relationship between the verbal and the visual—a relationship long-appreciated by the generations of storytellers, poets and writers who have conjured vivid, intricate imagery through the words they speak or write. Nevertheless, as much as text-to-image AI tools are poised to redefine humanity’s visual languages, Blind Camera reminds us that they do not represent the only way that AI can be harnessed to create visual art.

https://opensea.io/assets/ethereum/0xFe4e82E80D1e70c6B60eDE19A8BF0B63F5c1b4F3/122

Words, while adept at conveying multitudes, have never held a monopoly on defining our imagination or gaze. We have always relied on the full, multi-dimensionality of our sensory toolkit in our creative processes—synthesising the varied sights, textures, aromas and melodies around us into things that we want to see and to be seen. For example, Synchromism was a movement in painting in the early 20th century based on the idea that colours in a painting could be arranged in a similar manner as notes in a musical piece. Indeed, sound and other sensory inputs have rich visual associations, and Blind Camera can be appreciated as part of a longer tradition of artists tapping on these associations to create works of art.

In using AI, Blind Camera certainly stands out in terms of its methodology to express the relationship between sound and imagery. That said, it is not the use of AI per se that gives this work its conceptual heft.

What made me pause when looking at the images created by Blind Camera is realising that the series represented an entire system of perception—a novel way of seeing—from what we’re typically accustomed to. In Diego’s own words, each image is “indexical of the surrounding soundscape and not of the scene in front of the camera.” But I also think there is more to this: each image in the series was not the result of an individual’s gaze (or hearing) at a specific moment, the basis of a point-and-shoot camera, but an agglomeration of different sensory inputs that span across time and space. The works are pictures, yet more than pictures—not only because they were created from the soundscape of a particular place, but also because their expression was mediated through a synthesis of multiple sounds from multiple places, i.e. the training of its underlying AI model.

Chromesthesia may provide a useful frame of reference here. Also known as sound-to-colour synaesthesia, individuals with this condition can perceive colours, shapes and movement involuntarily when hearing sounds. Such perceptions, however, would be idiosyncratic—different individuals with the condition would likely experience the same sound differently. As someone without chromesthesia, I would also never be able to experience how someone with the condition could perceive sound, beyond what their own descriptions or expressions conveyed. The system of perception as represented by Blind Camera may thus be the best possible means for me to experience chromesthesia directly, and as a more objective phenomenon that goes beyond a particular individual’s experiences. In this sense then, the images of Blind Camera may reflect a collective form of chromesthesia—the aggregate of the myriad ways in which sound and imagery can be associated together.

https://opensea.io/assets/ethereum/0xFe4e82E80D1e70c6B60eDE19A8BF0B63F5c1b4F3/125

Beyond the conceptual novelty of Blind Camera, what has sustained my interest in the work over time is how it grounds me intimately in the local—the particular rhythms, textures and atmospheres that Mexico City encompasses, an “enormous improvised hypermetropolis” (per David Lida’s description of the city in his book, First Stop in the New World) heaving with almost ten million inhabitants and many more on its periphery.

The blurry, undeveloped yet evocative elements in the images spanning this series invite one to speculate on how the actual scene behind each piece would have looked like. One can almost imagine masses of people, vehicles, buildings within the images. Some seem to drown within the din of Mexico City, while others seem to find a space to emerge with clarity. Indeed, the limitations of this “blind” camera beckon and instigate us to look closer—to cast our own gaze towards the same subjects and try to illuminate them in our mind’s eye.

Naturally, I felt this impulse most strongly when I was physically in Mexico City, travelling there with my wife for a week last month. As we walked around different neighbourhoods within the city and took in its many sights (and sounds!), I couldn’t help but think about where Diego could have captured the sounds used to create the 100 images in the series—would my physical proximity to their underlying soundscapes allow me to see through the images?

Nevertheless, I soon realised that this train of thought was futile, for each image was not a reflection of the sensory landscape of the city but a refraction. Even though each of the 100 images was created based on the soundscape of a particular place at a particular moment in time, each of them should be more accurately seen as a composite built upon a larger set of sonic and visual data drawn from the city. The point of Blind Camera was not to recreate the sensory reality of specific localities across Mexico City, but to remix them through the specific medium of AI, taking in both its unique affordances and constraints.

In this regard, the artist’s hand in Blind Camera is most salient not in the decision on where to take the “snapshot” for each piece, but in what other data he has collected to train his AI model. After all, even though each image was seeded by the specific soundscape of a particular site, this sensory input would have been mediated by the “learnings” of the AI model gleaned from other sensory data from other sites within the city. That we can create an entire system of perception based on this process in the first place reveals the tremendous power of AI to find patterns and synthesise data. Yet, this also demonstrates the subjectivity of AI, for what was used to train the models ultimately defines the boundaries of their own “dark rooms”—the latent space from which they generate the things we prompt them to.

This subjectivity, however, should be seen as a feature and not a flaw for art created with AI. In the case of Blind Camera, we should not rely on it for an “objective” reimagination of Mexico City, for an individual’s experience of cities has tended to be highly personal. Art’s role in expressing or critiquing the character of a city is thus different from academia’s—not to tease out some objective truths from the writhing, burgeoning urban environments that increasingly define the contemporary human condition, but to marinate within its subjectivities and make us reflect on them in our own way. The use of AI simply injects another layer of subjectivity into this process.

That the Mexico City depicted through Blind Camera is only a subset of the city that the artist has seen and heard is thus to be expected. What is more crucial is whether the artist has left sufficient space for their works to transcend the subjectivity of their individual positions within the city.

On this, Blind Camera shares some similarities with Francis Alÿs’ Colector (1990-1992), in which the artist walked through the streets of Mexico City, pulling a small magnetic toy dog that collected various metallic debris along the way. Just as how the material output in Colector is dependent on Francis Alÿs’ walking path, the latent space of Blind Camera’s AI model is likewise bounded by the data of the city that Diego has collected. That said, in both works, we cannot see the full extent that the two artists have traversed the city spatially, beyond blurry snippets of where they might have been.

To fully appreciate the scope of their works then—the tangible products of their walks—we thus have to imagine the artists’ physical journeys ourselves based on our own impression of the city, whether we have been there physically before or not. What we make out of the detritus picked up by Colector or the images generated by Blind Camera will thus be shaped by our personal preconceptions and experiences of Mexico City as much as the artists’ own.

Indeed, Blind Camera provides a prism through which one can perceive Mexico City, where the local is refracted with the personal—the subjectivity of the artist melding into one’s own. In a way, it is a camera turned onto itself. What we see in the images it creates depends on all the prior sensory data that has gone into making this camera. This in turn prompts us to speculate on the artist’s journey through Mexico City, and inevitably leads us to consider our own relationship with the city.

For those who have never been to Mexico City, it will be worthwhile to keep Blind Camera in mind when you do have the opportunity to visit. The city can be incredibly overwhelming, inundated with multiple layers of history, complexity and nuance. In this respect, a work like Blind Camera, grounded in the intricacies of the city, will provide a handy accompaniment as you try to make sense of what you see and hear in this place that the Aztecs once conceived as the navel of the world.

https://opensea.io/assets/ethereum/0xFe4e82E80D1e70c6B60eDE19A8BF0B63F5c1b4F3/119

Credits: The header image of this essay was cropped from Blind Camera #4, one of the two pieces from this series that I have collected.