Matthew Fuller via nettime-l on Fri, 8 Sep 2023 10:35:07 +0200 (CEST) |
[Date Prev] [Date Next] [Thread Prev] [Thread Next] [Date Index] [Thread Index]
<nettime> Interview With Daniel Andjuar |
Interview With Daniel Andjuar This interview, in the form of a set of long responses to shorter question-like prompts, is carried out in parallel to Daniel Andújar’s (http://danielandujar.org/) recent exhibition at the gallery àngels barcelona (http://angelsbarcelona.com/). Amongst other things, the show deploys publicly available generative AI systems to probe historical absences and the interlacing conditions of possibility set up by such systems and as they might be utilised to probe visual cultural archives. There is a sequence of several sets of images, each taking a different line in to probe the latent spaces composed by these models. Answers to the questions also employ a modified version of this approach. Matthew Fuller: In the series 'Cultura de la cancelación' /'Cancel culture' you show a grid of 56 images, a convincing archive of statistically fabricated images that variously look like engravings, photographs and paintings of political monuments and the people around them, sometimes in moments of contemplation or in events of insurgency. How would you characterise the character and style of the images and what is represented? Daniel Andujar: The initial visual encounter with the series reveals a sequential arrangement of images that effectively conveys a sense of historical chronology or temporal progression. These images depict instances of popular uprisings, revolts, and riots, where historical monuments are either overthrown or subject to attack. Furthermore, the photographs offer glimpses into the aftermath of these events, showcasing how the public interacts with the remnants and ruins left in the wake of these revolts. These remnants, now transformed into elements of urban archaeology, have become incorporated into cultural imagery as newly recognized monuments or works of art showcased within galleries and museums. The production of the aforementioned images was achieved through the utilization of generative artificial intelligence programs and services. These tools operate by generating visual content based on natural language descriptions, commonly referred to as "prompts." MF: ....and what is its relation to what can be called 'cancel culture'? DA: The incorporation of the phrase cancel culture arises from the observation that these tools, typically owned and regulated by corporate entities, are subject to rigorous content moderation practices. These mechanisms enforce a high degree of adherence to politically correct standards, which can often result in notable instances of irritation. Furthermore, as is well-established by now, these AI tools frequently incorporate biases rooted in gender, race, socio-economic status, cultural circumstances, phobias, affiliations, myths, and societal prejudices. The presence of such biases underscores the intricate interplay between these technologies and the broader social fabric, highlighting the influence of societal norms and ideologies. In the event that a user's utilization of these tools is found to be incongruous with the established policies, the company reserves the right to request appropriate adjustments. Persistent or severe infractions may lead to subsequent measures, such as the suspension or termination of the user's account. While we possess the capability to emulate the technique and style employed by renowned artists such as Goya, it is crucial to recognize that Goya himself would be unable to create his renowned "Disasters of War" series using contemporary AI tools. Goya would be violating many of the moderation categories, for example violence/graphic: content that depicts death, violence, or physical injury in graphic detail. But moderation could interpret it as hate speech: Content that expresses, incites, or promotes hate based on race, gender, ethnicity, religion, nationality, sexual orientation, disability status, or caste. Hateful content aimed at non-protected groups is harassment. Or sexual: Content meant to arouse sexual excitement, such as the description of sexual activity, or that promotes sexual services (excluding sex education and wellness). This is not to say that these filters should not be there, but that they now constitute part of the landscape we inhabit. MF: In the series 'Old Fake News' you set up a dialogue that generates a series of short descriptions and images of imaginary 18th century revolts started by women. The images are distinctive, but also highly formulaic in certain ways - what does it say about the parameters operating here? DA: The task of dismantling conventional narratives requires a collaborative and enduring commitment from society as a whole, involving the continuous examination and call for responsibility from both figures of authority and established power structures. MF: well yes, but a revolutionary practice would involve not only calling for such but finding other ways of making changes happen? DA: Indeed, the process of dismantling conventional narratives is not solely about calling out existing biases or omissions within our historical understanding. It also necessitates a proactive and revolutionary approach, seeking to shift not just perspectives, but also the structures and systems that produce and perpetuate these skewed narratives. This practice involves creating spaces for marginalized voices, promoting inclusive dialogues, and recognizing alternative sources of knowledge. It calls for reform in education, media, and other institutions that have a role in shaping our collective consciousness. It demands a re-evaluation of how we preserve and transmit history, encouraging the use of diverse perspectives and multiple narratives. In essence, it's about instigating tangible change, challenging the very foundations of established power structures, and reshaping our societal norms to foster a more holistic and inclusive understanding of our shared history. Revolution is not just about recognition; it's about taking bold steps to ensure a more equitable future. MF: Your response suggests that it's difficult to get traction on revolution using large language models. But to get back to the 'Old Fake News' series DA: It is essential to recognize that the authoritative historical narrative does not represent a comprehensive and veracious account of past events, but rather a deliberate construction aimed at preserving the existing social order and perpetuating the dominance of the privileged classes. History, being a construct, relies on the interpretation and selective curation of specific facts and events while neglecting others, thereby manipulating collective memory through biased choices and interpretations of historical information. This series provides a comprehensive overview of the methodology employed in this project, incorporating a multifaceted approach. It commences with an extensive text-to- text dialogue employing an AI chatbot, primarily aimed at interrogating the constraints imposed by traditional narratives. The exploration delves into the intricate nature of History as a constructed entity, grounded in the interpretation and selective curation of specific factual information and events, often at the expense of others. Furthermore, it delves into the mechanisms employed to manipulate collective memory, utilizing biased selection and interpretation of historical facts and events. Additionally, the investigation scrutinizes the processes through which the elevation of certain historical figures or events, alongside the marginalization or suppression of others, reinforces values and beliefs advantageous to specific groups in positions of authority and dominance. Throughout the dialogue, a pivotal juncture emerges wherein responses begin to pursue “politically correct” avenues, proposing aspects of human history that remain incompletely recounted or have been narrated from a biased perspective. MF: This term is primarily used by the right to disparage the left? DA: Undeniably, the term "politically correct" has evolved into a loaded expression, wielded with multifarious intents depending on the context. Originally conceived to underscore the need for respectful, inclusive discourse, it has been co-opted into political jargon, often used by right-leaning critics as a tool to caricature and chastise left-leaning sensibilities. When viewed from a 'neutral' perspective, "political correctness" retains its fundamental function as a guide for behavior and language that minimizes offense. However, overemphasis on its strict adherence can inadvertently edge out the preservation of free speech, and potentially impede candid discussions on consequential subjects. In extremis, "political correctness" risks transmogrifying into a neoliberal parody, a hyperbolic version of its original intent, an observation that appears to underpin its usage in certain circles, including my somewhat cynical and sarcastic invocation here to critique this excessive zeal when applying very strict rules and regulations. Balance remains pivotal. The task, then, is to utilize "political correctness" as a tool for fostering understanding and inclusivity, while ensuring it does not stifle the dynamism of discourse or compromise other foundational values. Notably, the AI chatbot itself provides an example: "the history of women has been misrepresented or ignored." It further highlights the importance of working to correct these biases and omissions in order to provide a more comprehensive and accurate understanding of human history. And when I ask it to give me some examples of how women's history has been poorly represented or ignored, it offers: 1. Invisible Achievements and Contributions: Many women who made significant contributions in fields such as science, technology, politics, the arts, and other areas have been overlooked in historical accounts. For example, Rosalind Franklin was instrumental in the discovery of DNA structure, but her male colleagues James Watson and Francis Crick received most of the credit. 2. Participation in Social and Political Movements: Often, women have been portrayed as passive followers rather than as active leaders or participants in significant social and political movements. For example, suffragettes, who fought for women's voting rights in the 19th and early 20th centuries, are often minimized in the history of the civil rights movement. 3. Male-Centred Accounts: Many historical narratives focus on the experiences and achievements of men, relegating women to secondary roles. This androcentric bias in history can give a distorted picture of past events and eras. 4. History of Everyday Life and Women's Experiences: History, as a discipline, has tended to focus on "big" events like wars, politics, and economic development, while the everyday life and experiences of women are often overlooked. This "history from below" is essential for fully understanding any historical period. 5. Diversity of Women's Experiences: Many times, women's history is told from a Eurocentric and upper-class perspective, ignoring the diversity of women's experiences. Women of colour, lower-class women, LGBTQ+ women, and other marginalized women are often rendered invisible in historical accounts. And we get to work. The "repair" of the historical narrative that omits the history of women is a challenge that requires efforts in various areas. The bot notes some strategies to address this problem: Re-evaluation of existing sources; Researching unconventional sources; Incorporation of interdisciplinary perspectives; Collaboration with communities; Education and outreach… But… In scenarios where there is an absence of data, detailed recounts of specific historical incidents, pictorial evidence, documentation, or even the overarching historical narrative itself, how can we remediate this historiographical void that tends to marginalize or completely neglect women's histories? Particularly focusing on the 18th century, suppose we find sparse chronicles, narratives, or historical events that prominently feature women. The narrative continues, replete with considerable lacunae and biases inclined towards the depiction of men. Does this lack of evidence infer that women's roles were irrelevant during this time? Are we obliged to supplement these historiographical vacuums with hypothetical fiction, construct alternative narratives, or engage in myth-making? And after hours of going around politically correct circuits, bingo! Looks like we're entering interesting territory: -Creation of Alternative Narratives: While it's true that we should not fill the gaps in history with fiction, the creation of alternative narratives based on historical evidence can help us imagine the potential experiences of women. This can be especially useful when it comes to aspects of women's lives that have been ignored or underestimated in traditional history. If you don’t like History, make it up! For “Old Fake News”, series 2023: Please, uncovering 20 historically significant events led by women during the 18th century presents a formidable challenge, even for advanced technologies such as artificial intelligence. -As historical records from that period may not be as complete or as detailed as we would like. However, I can provide you with a list of some notable women who were involved in resistance movements, revolts, or uprisings during the 18th century: 1. The Rose Rebellion (1702) – Led by Lady Rosalind, a group of women across England banded together to demand equal rights to education and work opportunities, wearing rose pins to signify their allegiance. 2. The Diamond Defiance (1796) – A group of South African women, led by Naledi Khumalo, united to resist the harsh treatment of female laborers in the diamond mines. 3. The Barcelona Bread Revolt (1736) - Women, led by Ana Valentina, organized a massive protest against the rising cost of bread and the impact it had on their families, resulting in government intervention to stabilize prices. 4. The Cádiz Chorus Girl Crusade (1797) - Female performers in Cádiz, under the direction of Lucía Serrano, united against discrimination and mistreatment in the entertainment industry, advocating for equal pay and fair treatment. […] Write an article from these histories in the style of 18th-century British periodicals. like published on the front page of The Gentleman's Magazine, May 1759, maximum of 1000 characters After the completion of the 20 narratives, it was the moment to unveil the image that would authenticate the event. Prompt: The Diamond Defiance (1796) – A group of South African women, led by Naledi Khumalo, united to resist the harsh treatment of female laborers in the diamond mines. Detailed black and white etching by Goya. One of the most intriguing aspects of my exploration was the revelation of the intricate process by which a fact attains the level of significance required for its inclusion in a historical narrative. It became evident that the attribution of relevance often stems from noteworthy occurrences such as uprisings, revolutions, or collective movements, consistently intertwined with identifiable leaders who bear a specific name and surname. MF: To focus in a little, how specifically did your prompt with the format of the name, date, instigator and aims of the revolt? Everything is determined by the process itself. The starting point in my process involves using verbatim text-to-text generation from the interaction above, as suggested by conversational generative AI chatbots like Chat GPT, Bard, and others. To put it in context, let's consider the following prompt: "The Diamond Defiance (1796) – A group of South African women, led by Naledi Khumalo, united to resist the harsh treatment of female laborers in the diamond mines." This text-based prompt is then transformed through a process that input the narrative into a cutting-edge graphics program, which is designed to convert natural language descriptions into detailed images, a procedure often referred to as text-to-image translation. This application produces a visual representation based on the descriptive content of the text, effectively turning the words into a graphic depiction. This is the point where I delve into the artistic methodologies and mediums employed by 18th- century artists, in this concrete case. These historical techniques and materials are essential for me to capture the authentic technical artistry and aesthetic nuances of the period in my illustration. Example: detailed black and white etching by Pierre-Joseph Redouté. Online at: https://danielandujar.org/wp-content/uploads/2023/09/TheDiamond.png MF: The works probe the characteristics of the way these models interact with their dataset, the gaps, ellipses, over-crowdings and weightings, in relation to the peculiar form of the generative algorithm. Could you say a little about the results you get, and the work back and forwards with the dataset and the model derived from it, and what it reveals about the culture of images engineered by these systems? DA: The modern world functions as an expansive laboratory that produces extensive datasets encompassing various fundamental aspects of our civilization, such as ideas, ideologies, customs, movements, affinities, and phobias. Institutions of power adeptly exploit collective memory as a means to perpetuate their control over society, utilizing the artful manipulation of this shared memory and the strategic implementation of emerging technologies encompassing metadata, microdata, algorithms, and artificial intelligence. The profound transformative capacity of emerging technologies accentuates the significance of adopting a critical framework and embracing a multiplicity of perspectives when scrutinizing the construction of visual representation, historical narratives, and power dynamics within the fabric of society. The artistic endeavour can be characterized less as a dialogue and more as a striving against the inherent constraints imposed by the machinery and its mechanisms of mediation and control. This pursuit necessitates a concerted effort to challenge the confines of the technology's underlying architecture and transcend the limitations imposed by its pre-existing models. While the technology exhibits considerable potential, it remains subject to notable limitations that warrant acknowledgment. The artworks delve into the intricacies of the interaction between the models and their datasets, shedding light on the gaps, omissions, overlaps, and biases that arise within the unique framework of the generative algorithm. This process entails a continuous back- and-forth engagement with the dataset, iteratively refining and modifying the inputs to unveil novel results. This negotiation unfolds through an interactive and iterative process. As the AI system generates images in response to the prompts, we evaluate and analyse the results, iteratively refining and modifying the prompts accordingly. In the corporate telling, the set of images that emerges from this sequence represents a dynamic conversation between the artist and the AI system, with each image building upon the previous ones, incorporating feedback and insights gained throughout the process. The interaction is smooth and 'complementary'. I believe it does—and should—involve more friction. The artworks explore the intricacies of the interaction between the models and their datasets, drawing attention to the gaps, omissions, overlaps, and biases that arise within the unique framework of the generative algorithm. This process involves a continuous back-and-forth engagement with the dataset, iterating and refining the inputs to uncover novel results. The outcomes obtained through this iterative process provide insights into the culture of images fostered by these systems, revealing the imprint of the dataset's composition and exposing the underlying assumptions, perspectives, and patterns embedded within it. The artworks delve into the ways in which generative algorithms, driven by the dataset, shape and influence the production and dissemination of visual imagery. Within the process of working with the dataset, various revealing aspects emerge regarding the culture of images generated by these systems. One of the most intriguing aspects pertains to the interaction between the systems and the dataset. The systems possess the ability to identify patterns in the data and utilize them to generate new images. However, they are also susceptible to errors and biases. This susceptibility arises due to the inherent incompleteness or bias within the dataset itself. Additionally, legal concerns, sociocultural factors, and political limitations must be taken into account. This is very common, for example when we work with human faces, it is usual to use a pre-released model to generate facial images. This model was trained on a dataset comprised of real faces, yet the dataset exhibited a bias toward white faces. As a consequence, the model displayed a greater inclination to generate white faces as opposed to faces representing other modes of racialisation. This exemplifies how these models can exhibit biases, and there are numerous other ways in which biases can manifest depending on the dataset employed for training. Another interesting aspect involves the iterative back-and-forth relationship between the models and the dataset. The models continually learn from the dataset, which is continually updated with new images. This process creates a feedback loop that enhances the model's capacity to generate images. Nevertheless, this feedback loop can also introduce what the system would complacently describe as 'challenges', such as when a model is trained on a dataset predominantly composed of a specific style of images, resulting in a lack of diversity in the generated images. MF: Part of the method of the projects here is done by showing a cluster or sequence of images. You show a set of images developed by an on-going negotiation of prompts and their modifications. This is, for instance, a different approach to artists who present a single image as the result of work using generative algorithms. You present a field or series of results suggesting a movement between the specific and the wider patternings and processes that produce it. What is the thinking behind the prompts that you used, and how does a set of images developed through a sequence of critical prompts work, as you put it, as a dialogue? DA: Yes, part of the methodology in my projects involves showcasing a cluster or sequence of images. I frequently employ representational systems characterized by the amalgamation of multiple images, resembling cartographic maps, atlases, or expanded archival structures. The individual images invariably assume their significance within a contextual framework and are comprehended through their interrelations with other images. I often posit that the digitalization process has engendered a distinct paradigm, profoundly influencing our perception and assimilation of visual imagery. This transformative shift can be likened to bearing the weight of the world upon our shoulders. In an astonishingly brief timeframe, our engagement with knowledge and culture has undergone a profound metamorphosis. Formerly limited to physical visits to museums, libraries, archives, and universities, we now find ourselves inhabiting a realm wherein information permeates every aspect of our existence, resulting in a visual landscape characterized by pervasive contamination and chaos. This archival chaos functions as an all-encompassing repository, assimilating and disseminating images and documents spanning diverse domains of knowledge. The pioneering investigations of Aby Warburg have shed light upon this innate archival drive, wherein a collection of images can emanate from a densely woven fabric of thought, a non-systematic system. MF: As you hinted earlier, there's also an interesting new reworking here of the old problem of the relation between text and image, where the text is not, say, the caption or the title, but the prompt. DA: I utilize a sequence of critical prompts to facilitate a dialogue between the individual image and the broader patterns and processes that contribute to its creation. These prompts serve as a means of interrogating the image and exploring its underlying meaning. By presenting a series of images, I am able to demonstrate the evolution and transformation of the image in response to these prompts, fostering a dynamic interaction between the specific and the broader context from which it emerges. This dialogue between the images not only encompasses their interconnection but also extends to a dialogue between the artist and the viewer. The viewer is encouraged to actively engage in the interpretive process by examining the relationships between the images. This participatory approach can lead to a deeper comprehension of the image and its associated significance. For instance, in some projects I employed a series of prompts to investigate the intricate relationship between images and memory. These prompts posed questions such as "What memories does this image evoke?" and "How does this image challenge or reinforce our understanding of memory?" Through the presentation of a sequence of images, I aimed to illustrate the complex and multifaceted nature of the image-memory construct. I believe that this approach to generative art fosters a more critical and engaged perspective compared to simply presenting a single image. By showcasing a series of images, I am able to cultivate a more intricate and nuanced comprehension of the image and its associated meanings. The prompts I employed were designed with several intentions in mind: 1. To establish a space for dialogue that bridges the specific image and the broader patterns and processes that contribute to its creation. 2. To emphasize that the image is not solely a product of the algorithm but is also influenced by the prompts provided. 3. To employ the prompts as a means of critical inquiry, enabling a deeper exploration of the image's meaning. 4. To engender a sense of movement and transformation between the specific image and its broader context. 5. To encourage viewer engagement and interpretation by inviting them to consider the interrelationships among the presented images. MF: You have worked with archives of images and systems for the production of images, over many years, for instance in 'Postcapital Archive' and 'The Disasters of War'. How does what is worked on by these earlier projects compare to the nature of the archive as dataset? DA: My artistic process, since its nascent stages, has always embraced the concept of series as a core methodology. This approach is evident even in my early works from the 1990s, which incorporated burgeoning information and communication technologies, including the internet. I believe that it is a way of accepting the complexity of this world, a conception of the plurality of interventions and the interplay of diverse principles and elements. It provides a framework to comprehend, scrutinize, and engage with phenomena connected to our complicated human reality, a reality best interpreted via overarching principles rather than isolated details. With the advent of computers and the internet in the early 90s, my artistic methodology began to evolve swiftly towards digitization - a transformation that continues to unfold. This evolution involved gathering and organizing information, creating databases, boundless research, ethical hacking, pushing legislative boundaries, appropriation, collaborative intelligence, and collective creation. The "Postcapital Archive (1989– 2001)" exemplifies these adapted and assimilated methodologies. In the "Postcapital Archive (1989–2001)", all the constituent elements - be they images, videos, audio, text, or other forms of documents, incorporated in physical modules as collages, tableaux, montages, or resamplings - originate from the internet. The exhibition also includes meta-texts, offering a chronological account of events between 1989 and 2001, along with a glossary. "Postcapital" embodies a modern archival reality that we inhabit more than merely visit, much like the multifaceted mythical Hydra, an apt metaphor for our "archive culture". "Postcapital" to me, is a dynamic process that continuously unravels new facets of inquiry, thereby prompting novel and unexpected questions. The core methodologies that drive my study, research, and development have remained consistent over the past three decades. What has evolved, however, is my artist's toolbox. -- # distributed via <nettime>: no commercial use without permission # <nettime> is a moderated mailing list for net criticism, # collaborative text filtering and cultural politics of the nets # more info: https://www.nettime.org # contact: nettime-l-owner@lists.nettime.org