Skip to main content

Artificial Intelligence, Art and Indigeneity: 9. Desde una cuenta propia hasta un modelo propio From an account of one’s own to a model of one’s own

Artificial Intelligence, Art and Indigeneity
9. Desde una cuenta propia hasta un modelo propio From an account of one’s own to a model of one’s own
  • Show the following:

    Annotations
    Resources
  • Adjust appearance:

    Font
    Font style
    Color Scheme
    Light
    Dark
    Annotation contrast
    Low
    High
    Margins
  • Search within:
    • My Notes + Comments
    • Notifications
    • Privacy
  • Project HomeArtificial Intelligence, Art and Indigeneity
  • Projects
  • Learn more about Manifold

Notes

table of contents
  1. Praise Page
  2. Understanding Languages, Cultures and Societies
  3. Title Page
  4. Copyright
  5. Dedication
  6. Contents
  7. List of figures
  8. Notes on contributors
  9. Acknowledgements
  10. Introduction
    1. Notes
  11. Part I: AI, Art and Indigeneity
    1. 1. AI and art
      1. Notes
    2. 2. Indigenous AI
      1. Notes
    3. 3. Indigenous art and AI
      1. Notes
    4. 4. Indigenous AI and art in a Latin American context
      1. Notes
    5. 5. A timeline of AIAI
      1. Notes
    6. 6. Methodology
      1. Notes
  12. Part II: Between Dreams and Hallucinations
    1. 7. Autenticidad creativa y sensibilidad visual artificial Creative authenticity and artificial visual sensibility
      1. Notes
    2. 8. Inteligencia Artificial o los artificios de la inteligencia Artificial Intelligence or the artifices of intelligence
      1. Notes
    3. 9. Desde una cuenta propia hasta un modelo propio From an account of one’s own to a model of one’s own
      1. Una cuenta propia
      2. Un modelo propio
      3. An account of one’s own
      4. A model of one’s own
    4. 10. A Inteligência Artificial e a cosmofloresta, e Representando os parentes e os encantados com IA Artificial Intelligence and the cosmoforest, and Representing Indigenous kin and our spiritual beings with AI
      1. A Inteligência Artificial e a cosmofloresta
        1. Cosmofloresta: uma visão envolvente da natureza
        2. A Inteligência Artificial ao serviço da cosmofloresta
        3. Geração de imagens por algoritmos: uma ponte visual
      2. Representando os parentes e os encantados com IA
      3. AI and the cosmoforest
        1. The cosmoforest: an enveloping vision of nature
        2. Artificial Intelligence in the service of the cosmoforest
        3. Algorithmic image generation: a visual bridge
      4. Representing Indigenous kin and our spiritual beings with AI
      5. Notes
    5. 11. Hasta sueño con la IA I even dream about AI
      1. Notes
    6. 12. Nikiékliwahi e o espírito da máquina, e Realismo digital indígena Nikiékliwahi and the spirit of the machine, and Indigenous digital realism
      1. Nikiékliwahi e o espírito da máquina
      2. Realismo digital indígena
      3. Nikiékliwahi and the spirit of the machine
      4. Indigenous digital realism
      5. Notes
  13. Part III: Generative Thoughts
    1. 13. A genealogy of stereotypes: the representation of Indigenous Peoples with generative AI
      1. Mise-en-scène
      2. The pose and the screen
      3. The ethnic scene
      4. Under-representation
      5. Notes
    2. 14. Drawing the line: creativity, censorship and copyright in Indigenous engagement with AI image-generation tools
      1. Creativity
      2. Censorship
      3. Copyright
      4. Notes
    3. 15. Dragon dreaming: notes on methodology for the collaborative design of an Indigenous image-generation prototype
      1. Leeds workshop 1: broken
      2. The past is ahead of us (field note)
      3. Commission for the prototype
      4. Methodology: to hold a space
      5. Leeds workshop 2: risk and ritual
      6. Holding your words in my mouth (field note)
      7. Leeds workshop 3: What do you mean by ‘intelligence’?
      8. Listening through my eyes (field note)
      9. Leeds workshop 4: What are your dreams?
      10. A model and a dog
      11. Prisoners of science (field note)
      12. Leeds city market and the global circulation of images of Indigeneity
      13. Indigenous data security
      14. Final thoughts: AI and Indigenous epistemologies
      15. Notes
    4. 16. Algorithmic Indigeneity, Indigenous textiles and future imaginaries
      1. Algorithmic logic in Indigenous cultures
      2. Indigenous textiles in AI
      3. Indigenous future imaginaries in AI
      4. Notes
    5. 17. Artificial Intelligence, appropriation, ancestrality and Buen Vivir
      1. Indigenous Appropriation of Artificial Intelligence (Sebastián Gerlic)
      2. Ancestral Intelligence (Sebastián Gerlic)
      3. Digital Buen Vivir (Alex Potiguara)
      4. Notes
    6. 18. Is it art? Exhibiting the AIAI project
      1. The Generative Fictions Exhibition: Words, Images and Territories (Sandra De Berduccy)
      2. Science Week at the National Botanic Gardens of Ireland (Andreas Rauh)
      3. Notes
    7. Conclusion
      1. Epistemologies, creativity and authorship
      2. Languages and translations
      3. Data sovereignty and cultural stewardship
      4. Environmental (un)sustainability and the limits of generative AI
      5. AI (counter)hegemony and the limits of decolonising Indigenous AI
      6. Ancestrality and AI
  14. Bibliography
  15. Index

Chapter 9 Desde una cuenta propia hasta un modelo propio From an account of one’s own to a model of one’s own

Haylly Zamora Aray

Una cuenta propia

Soy artista visual, artesana e investigadora indígena, miembro de los Pueblos Wichí del norte argentino y Kariñ’a del estado Bolívar de Venezuela. En este momento resido en el Chaco, Argentina. Mi trabajo artístico fusiona técnicas tradicionales con innovaciones digitales desde un enfoque colaborativo y de diálogo intercultural. Desde el 2022, colaboro con la serie de libros Abya Yala con Amor (de la ONG Thydêwá) donde conocí a Sebastián, quien hizo de puente para conocer a Thea y así participar en el proyecto AIAI.

Al oír del proyecto AIAI yo no tenía mucha información sobre las IAs y aún tengo poco conocimiento. Sin embargo, el acercamiento a estas (las generativas de imágenes) me trajo mucha curiosidad y ganas de seguir creando prompts e imágenes en mis tiempos libres; también de pensar y encontrar formas en que estas herramientas tecnológicas de generación de imágenes puedan servir en procesos de educación.

Así que mi primer paso fue crear mi propia cuenta de Midjourney (me ayudó mi sobrino) en donde experimentamos mi familia y yo muchas formas de crear los prompts. Ensayamos con prompts sencillos como ‘una mujer bella’ escrito en diferentes idiomas, para ver si el idioma dictamina el color de la mujer. (Todas salieron blancas y occidentales, a menos que añadiéramos una palabra como ‘indígena’, y las indígenas siempre se parecían a las de México y no a nosotros, del Chaco argentino, y si poníamos ‘native’ en vez de ‘indigenous’ en el prompt en inglés, salían como nativas americanas). Y ensayamos también con otros prompts para ver si podíamos llegar a las imágenes más cercanas a lo que queríamos o teníamos en nuestra mente.

En el 2024, a través del Programa de Cultura para los Pueblos Originarios del Gobierno del Chaco, algunos maestros indígenas y yo hicimos una ronda de talleres llamados ‘wela/nala’ (luna/sol, en Wichí Ihämtes) en donde hablamos de la generación de imágenes con IA y la representación de gente que se parece a nosotros, originarios del Chaco, además de sus usos posibles. Empezamos pidiendo a los chicos describir en palabras a una persona del Chaco, o a un amigo local de ellos y esto les resultó muy complicado, así que generar imágenes con IA que se parecen a nuestro pueblo nos parece sumamente difícil de conseguir a través de prompts textuales. Sin embargo, otra idea que se nos presentó fue cómo, por medio de esta experiencia de generar imágenes y criticar los resultados con los chicos, se podrían generar también diálogos entre pequeños grupos de jóvenes estudiantes de secundaria y docentes indígenas.

Un paso más personal fue el subir imágenes de mis obras y jugar con ellas dentro de la IA y ver como esta las leía y qué tipo de imágenes me devolvía. A veces subía una foto de una obra mía y pedía que la IA generara un prompt textual en inglés a base de esta obra. Luego utilizaba este mismo prompt para generar nuevas imágenes. También, mientras experimentaba con Midjourney, jugué con otras IAs para ver la calidad de las imágenes y qué tan accesibles eran para las personas. Luego Meta sacó su IA con WhatsApp, que es la aplicación de mensajes que usan casi todos en Argentina, y esto hizo que esa brecha se achicara.

En ese momento, cuando tenía tantas imágenes, pensé que tenía que hacer algo con ellas. Por eso abrí un Instagram para subir algunas y hacer comparaciones de las imágenes hechas por mí y las creadas por la IA. Y desde ese momento también reflexioné sobre si realmente se necesita tener tantas imágenes que no serán usadas: ¿Qué pasa con esas imágenes? ¿Qué podemos hacer? ¿Podemos hacer reciclaje digital? (Estoy siempre llena de preguntas.…)

El Hombre puma proviene de los relatos ancestrales del pueblo Wichí. Las imágenes generadas con IA (véase Fig. 9.1, imágenes del centro y de abajo) se basan en una descripción textual de una pintura que había hecho antes (véase Fig. 9.1, imagen de arriba). Usé el traductor de Google para lograr hacer el prompt en inglés. La imagen del centro es la que resultó más próxima a la que tenía en mi mente por los grafismos y la forma zoomorfa que tiene su cara que, para mi imaginario, está equilibrada y podría ser real.

La Mujer con wiphala es una imagen inspirada en una pintura mía (véase Fig. 9.2, imagen de arriba) hecha a partir del sentimiento de las luchas indígenas por preservar nuestro territorio. La del centro, de colores más opacos, fue recreado con Stable Diffusion a base de una foto de la pintura misma, sin un prompt textual. La de abajo, de colores más fuertes, la hice con Midjourney usando un prompt textual en español junto con una foto de la obra y coloqué ‘Fuji film’ en el prompt para que me diera una imagen realista de colores fuertes. Me gusta el resultado en este caso y me parece más próximo a mi obra, pero la mujer tiene pinta de ser una indígena mexicana y la IA claramente no ‘entiende’ lo que es una wiphala ni el significado de la distribución de los cuadritos de colores particulares.

Un modelo propio

Me ha gustado poder estar en cada paso de este proyecto porque vamos avanzando juntos y a nuestro ritmo y tomando en cuenta las necesidades del momento de cada artista. En esta nueva etapa con Stable Diffusion he podido experimentar la IA desde otra perspectiva, donde hay un poco más de control en lo que generamos. Para mí ha sido un poco más difícil porque es otro nivel de complejidad. Sin embargo, una vez entendido y superado el proceso de entrenamiento de la IA y al llegar a la generación de las imágenes, se vuelve más divertido el imaginar junto con la máquina. Yo la entrené con los tejidos tradicionales Wichí, pensando en un dragón porque así se llama la laptop donde está guardado nuestro prototipo (véase Fig. 9.3).

Me impresionó que algunas de las imágenes que me regaló el prototipo fueran muy parecidas a artesanías que son hechas en las comunidades, y le mostré a mi papá y me preguntó quién las vendía. Le dije que eran hechas por Inteligencia Artificial, y él me contestó que yo podía hacerlas. Yo no sé hacer el tejido tradicional, pero me hizo pensar en eso, en que podía aprender a tejer. Aunque siempre pensé que quería hacerlo nunca me había dado el tiempo hasta ver esta imagen de IA. Así que he comenzado a tejer.

Top image: Oil painting of a humanoid figure with the head of a black panther, shown in profile against a bright orange background with yellow teardrop-like patterns. The figure has a muscular torso and wears a necklace of red triangular shapes and a patterned yellow-and-black garment. He holds a wooden staff.

Middle image: AI-generated photograph of a head-and-shoulders portrait, facing straight to camera, of a figure with prominent cat-like ears and an upper body covered in intricate geometric tattoos in black-and-brown tones. The symmetrical patterns span the chest, shoulders and arms, creating a striking design reminiscent of Indigenous body-painting. The background is solid black, emphasising the details of the figure.

Bottom image: AI-generated photograph of a sleek, feline-like creature sitting on a rocky surface in an outdoor setting, facing to the left, looking out over a desert landscape in the background. Its body is covered in intricate, symmetrical patterns in dark-and-light tones reminiscent of Indigenous body-painting, creating a striking contrast against its natural form. The lighting emphasises the contours and detailed markings.

Fig. 9.1. Haylly Zamora Aray, Hombre puma (arriba: pintura original; abajo: IA) / Puma Man (top: original painting; below: AI). Details on page 72.

Fig. 9.1 details. Haylly Zamora Aray, arriba: Hombre puma, pintura al óleo, junio de 2023; centro y abajo: imágenes del Hombre puma generadas con Midjourney v.5.2, junio de 2023. Prompt original en inglés: ‘Un puma con forma humana, de estilo indígena de Argentina, con dibujos en la piel’.

Haylly Zamora Aray, top: Puma Man, oil painting, June 2023; centre and bottom: images of Puma Man generated with Midjourney v.5.2, June 2023. Prompt: ‘A human-shaped puma, with an indigenous style from Argentina, with drawings on the skin’.

Top image: Colourful watercolour painting featuring the head and shoulders of a dark-skinned woman with long blue hair, eyes closed with geometric red-and-black patterns painted on her torso. The background consists of a vibrant checkerboard of multicoloured squares – green, yellow, orange, pink and blue – with bold red paint drips running down across the grid.

Middle image: AI-generated photograph of a dark-skinned woman with long black hair and small gold dangly earrings, shown head and shoulders, eyes half-closed and splatters of red-and-white paint on her face. Behind her, out of focus, there seems to be a chequered cloth with beige, terracotta and French blue tones.

Bottom image: AI-generated photograph of an Indigenous woman shown head and shoulders, looking slightly away from the camera, with long black hair and wearing vibrant traditional clothing with intricate multicoloured patterns, including reds, blues and greens. Large, colourful beaded earrings hang prominently, and splodges of red paint drip down her face. Behind the figure is a hand-painted textile featuring a grid of bright squares in blue, orange, yellow and red. The background shows blurred market stalls with piles of fruit and vivid decorations.

Fig. 9.2. Haylly Zamora Aray, Mujer con wiphala (arriba: pintura original; abajo: IA) / Woman with Wiphala (top: original painting; below: AI). Details on page 73.

Fig. 9.2 details. Haylly Zamora Aray, arriba: Mujer con wiphala, pintura acrílica sobre papel, julio de 2023; centro: imagen de Mujer con wiphala generada con Stable Diffusion, julio de 2023; abajo: imagen de Mujer con wiphala generada con Midjourney v.6.0, marzo de 2024. Prompt original: ‘Mujer indígena con la wiphala de fondo. Fuji film [película Fuji]’.

Haylly Zamora Aray, top: Woman with Wiphala, acrylic painting on paper, July 2023; centre: image of Woman with Wiphala generated with Stable Diffusion, July 2023; bottom: image of Woman with Wiphala generated with Midjourney v.6.0, March 2024. Original prompt in Spanish and English: ‘Indigenous woman with the wiphala in the background. Fuji film’.

Set of four images. The top two are photographs showing traditional loose-woven plant-fibre textiles with geometric patterns in brown, beige and dark blue. The top-left image shows a woman holding the textile in front of her with adobe house, dusty yard and trees in background; top right is a close-up of the textile showing the detail. The bottom-left image shows an AI-generated model of a dragon made with beige fibres, against a plain dark background. The bottom-right image shows a model of a small animal, possibly a fox, made with a range of earthy-coloured fibres, against a plain white background.

Fig. 9.3. Haylly Zamora Aray, arriba, izquierda y derecha: fotografías de textiles Wichí utilizadas para entrenar su modelo LoRA; abajo, izquierda: Dragón Wichí, imagen generada con el prototipo IndigenIA utilizando su propio modelo LoRA basado en los textiles Wichí; abajo, derecha: fotografía de artesanía tradicional Wichí, elaborada con fibra de chaguar (Bromelia hieronymi), para mostrar la similitud con la imagen generada, junio de 2025.

Haylly Zamora Aray, top left and right: photos of Wichí textiles used to train her own LoRA model; bottom left: Wichí Dragon, image generated with the IndigenIA prototype using her own LoRA model based on Wichí textiles; bottom right: photo of traditional Wichí craftwork, made with chaguar fibre (Bromelia hieronymi), to show the similarity to the generated image, June 2025.

An account of one’s own

I am an Indigenous visual artist, artisan and researcher, a member of the Wichí people of northern Argentina and the Kariñ’a people of Bolívar state, Venezuela. I currently reside in Chaco province, Argentina. My artistic work fuses traditional techniques with digital innovations through a collaborative approach and intercultural dialogue. Since 2022, I have been collaborating on the Abya Yala con Amor book series (run by the NGO Thydewá), where I met Sebastián, who served as a bridge between me and Thea, and that’s how I came to participate in the AIAI project.

When I heard about the AIAI project, I didn’t have much information about AI models, and I still don’t really know that much about them. However, starting to explore these tools for generating images sparked a great deal of curiosity in me and a desire to continue creating prompts and images in my free time, as well as to think about and find ways in which these technological tools for image generation can be used in educational processes.

So, my first step was to create my own Midjourney account (with the help of my nephew), where my family and I experimented with many ways of creating prompts. We tried simple prompts like ‘a beautiful woman’, written in different languages, to see if the language dictates the woman’s skin colour. (They all came out white and Western unless we added a word like ‘Indigenous’, and the Indigenous ones always looked like Indigenous people from Mexico and not like us, from the Argentine Chaco. And if we put ‘native’ instead of ‘Indigenous’ in the English prompt, they came out as Native Americans.) We also tried other prompts to see if we could come up with images that were closer to what we wanted or had in mind.

In 2024, some Indigenous teachers and I conducted a series of workshops through the Chaco government’s Cultural Programme for Indigenous Peoples called ‘wela/nala’ [moon/sun in Wichí Ihämtes]. We discussed image generation with AI tools and the representation of people who resemble us, Indigenous Peoples of the Chaco, as well as the potential uses of AI tools. We began by asking the students to describe in words a person from the Chaco, or a local friend of theirs, and they found this very difficult. Generating images with AI that resemble our people seems extremely difficult to achieve through textual prompts. However, another idea that we came up with was how, through this experience of generating images and critiquing the results with the young people, dialogues could also be generated between small groups of high school students and Indigenous teachers.

A more personal step was uploading images of my own artworks and playing with them within the AI tool, seeing how it read them and what kind of images it gave me in return. Sometimes I would upload a photo of one of my works and ask the AI to generate a text prompt in English based on that work. I would then use this same prompt to generate new images. Also, alongside experimenting with Midjourney, I played with other AI tools to see the quality of the images and how accessible they were. Then Meta launched its AI tool within WhatsApp, the messaging app used by almost everyone in Argentina, and this narrowed the gap in terms of who could access these tools.

By that point, I had so many images that I thought I really needed to do something with them. So, I started an Instagram account to upload some of the images and compare the images I took of my own artworks with those created by AI. And from that moment on, I also started asking myself whether it’s really necessary to have so many images that won’t be used: What happens to those images? What can we do about all this waste? Can we do virtual digital recycling? (I’m always full of questions.…)

The Puma Man comes from the ancestral stories of the Wichí people. The AI-generated images (Fig. 9.1, centre and bottom images) are based on a textual description of a painting I had previously made (Fig. 9.1, top). I used Google Translate to create the prompt in English. The image in the middle is the closest to what I had in my mind due to its graphics and the zoomorphic shape of the face, which, in my imagination, is balanced and could be real.

The Woman with Wiphala is an image based on a painting of mine (Fig. 9.2, top), inspired by my feelings in relation to Indigenous struggles to preserve our territory. The image in the middle with the duller colours was created with Stable Diffusion based on a photo of the painting itself, without a text prompt. The image with the more vibrant colours on the right was created with Midjourney with a text prompt in Spanish along with a photo of the artwork. I added ‘Fuji film’ to the prompt to give me a realistic image with strong colours. I like the result in this case, and it seems closer to my work, but the woman looks like an Indigenous woman from Mexico, and the AI clearly doesn’t ‘understand’ what a wiphala flag is or the meaning of the arrangement of the coloured squares on it.

A model of one’s own

I’ve enjoyed being a part of every step of this project because we’re moving forward together, at our own pace, taking into account the current needs of each artist. In this new phase with Stable Diffusion, I’ve been able to experience AI from a different perspective, where there’s a bit more control over what we generate. For me, it’s been a bit more difficult because it’s a different level of complexity. However, once you’ve understood and mastered the AI training process and got to the point of generating the images, imagining and cocreating with the machine becomes more fun. I trained it with traditional Wichí textiles, thinking of a dragon because that’s what the laptop where our prototype is stored is called (see Fig. 9.3).

I was impressed that some of the images that the prototype gave me were very similar to handicrafts made in our communities, and I showed them to my dad, and he asked who sold them. I told him they were made by Artificial Intelligence, and he replied that I could make them myself. I don’t know how to do traditional weaving, but it made me think that I could learn. Although I always thought I wanted to do it, I’d never got around to it until I saw this AI image. So now I’ve started weaving.

Annotate

Next Chapter
10. A Inteligência Artificial e a cosmofloresta, e Representando os parentes e os encantados com IA Artificial Intelligence and the cosmoforest, and Representing Indigenous kin and our spiritual beings with AI
PreviousNext
This work is licensed under a Creative Commons Attribution-Noncommercial-Nonderivatives 4.0 International (CC BY-NC-ND 4.0) license.
Powered by Manifold Scholarship. Learn more at
Opens in new tab or windowmanifoldapp.org