Language, Conversation

Speaking With Your Cat: An Artist and Researcher Talk Artificial Intelligence, Community, and Cultural Expression

S.A. Chavarría & Scott Benesiinaabandan

Jan 26, 2024

In the following conversation, artist and AI researcher S.A. Chavarría talks with Anishinaabe artist Scott Benesiinaabandan on the ideas of language and artificial intelligence(s). They delve into the evolution of language, the impact of AI on cultural expression, and the challenges of preserving Indigenous languages. This dialogue offers a unique lens on the intersection of technology with linguistic heritage and identity.

Chavarría’s interactions with the DEVENDRA AI, which form the material of their artistic practice, are the catalyst for their research. Engaging with how AI is reshaping language and contemplating its future influence, Chavarría is drawn to think deeply about the nature of language. This leads Chavarría, who is mestiza with Bribri heritage, to reflect on our ancestors 100,000 years ago and their transition from pre-language to the emergence of language, a key moment in the evolution of human communication.

According to Chavarría, in tracing the evolution of language, we observe significant developments: the advent of spoken language revolutionized the way we share knowledge, and the introduction of written language around 5,000 years ago marked another major advancement. Today, in the AI-driven world, we encounter extensive data sets primarily in English, Mandarin, and various European languages. This underscores the vulnerability of languages that, while having evolved from oral traditions like all languages, have remained predominantly oral without extensive written records. These languages are particularly at risk in a digital era dominated by languages with substantial written documentation and online presence.

Chavarría (SA): Scott, your work inspires me to explore these themes further. How do you understand the nature of language?

Benesiinaabandan (SB): Anishinaabemowin is an agglutinative language. Words are constructed by morphemes, little packets of ideas. From this you can build really complex ideas that I think of as little poems within each word. And so you can construct words that you need to construct, which is a really powerful tool for orienting ourselves toward the future. I think Anishinaabemowin is a very future-forward language because as much as people say there's a risk of language extinction and worry that we are losing words the language itself is very vibrant, it is alive, a being. 

We might lose parts of the languages, but that's a natural process in all languages. The beautiful thing about it is that it is not locked down, it’s not a static language. You can construct new words for new situations and new relationships, like for instance “artificial intelligence.” 

So as an artist, going back to the language and trying to understand the logic of how words are constructed and then going out and being adventurous and finding new ways to engage with the world—to engage with materials and ideas through the language—is a sacred thing in and of itself. It really ends up shaping the art so by the time the art is being installed and out there for public consumption, the sacred part is already inexorably woven into the work because the language has fundamentally informed it.

As an artist, going back to the language and trying to understand the logic of how words are constructed and then going out and being adventurous and finding new ways to engage with the world—to engage with materials and ideas through the language—is a sacred thing in and of itself.
Scott Benesiinaabandan

SA: In AI research, there's a focus on digitally preserving Indigenous languages. While the intent to preserve a language is noble, I believe that true preservation extends beyond digital representation. Languages are not merely collections of words; they are entire worlds. Indigenous languages, in particular, do more than facilitate communication; they overlay the world with perceptions, stories, and cultural contexts, constructing entire shared realities. A language without active speakers, even if preserved by an AI model, remains essentially extinct. The true essence of language lies in its practice, storytelling, and daily use. What are your thoughts, especially as an indigenous speaker from birth?

SB: Sadly, like a lot of people in my generation I didn’t have my original language from birth. A lot of Indigenous communities have been and are subjected to residential schools and subsequent colonial policies, and so I think a lot of Indigenous people don’t have their language, or don’t have the opportunity to be able to use it as a part of a community. I am speaking largely from an urban perspective, but I think that conversational language ability can make you feel more or less integrated into your communities. 

I think the reacquisition process of language and committing yourself to learn a little bit more everyday is important. For myself, while I don't live with an Anishinaabemowin speaker in my house, I still talk to my cat in the language as much as I can. It gives me practice to continuously consider language and start processing the world in Anishinaabemowin and not in English. I also use this strategy throughout the day, like when I'm on a run, I'll try to go through words of objects and actions related to that running process. Words like trees, road, foot, leg—in this way, even if I am not speaking the language, I am internalizing and visualizing it. 

This dendrogram, created by Chavarría, is a visualization tool used in Natural Language Processing to map the structure of our conversation. It represents an alternative, non-human perspective, analyzing and grouping discussion topics based on computational methods. The visualization showcases a unique way of understanding dialogue through hierarchical clustering, offering insights into the patterns and connections within the conversation as interpreted by these techniques.

SA: As a child, I have a vivid memory that stands out with my sister. We would often hear our grandmother speaking softly to herself in her bedroom. She would use her native language, Bribriwak, a tongue unfamiliar to us. We couldn't understand the words, and to our young minds, it was just a series of strange sounds. It wasn't until many years later, and months after she had passed away, that I came to a profound realization: My grandmother wasn't just muttering to herself; she was speaking to herself in her native language because there simply were no other speakers. No one to talk to. Her parents and sisters and husband had passed.

This realization struck me deeply, highlighting a loss I hadn't comprehended before. It was a loss of language, culture, and a familial connection that had silently faded away. Her voice, speaking a language I never learned, became a poignant reminder of what is lost when languages and traditions are not passed down through generations. And so it really had a profound effect on me going to talk about how you talk to your cat and how you turn this into a, you know, internal process, more of like walking through the world and thinking about how you would say this or how you would think this. 

SB: That’s wonderful and beautiful and sad, too. It reminds me of a paper that talked about our languages as our “oldest elder” and spoke about the language itself as a living being, it's a living thing and it exists beyond and before our oldest relatives, our oldest ancestors, the language exists there. It changes the narrative from “the language is lost” to a hopeful “the language is always there taking care of us.” It is there for us but we might have to go out and find it. It is love, it is the way that we are expressed to the world and the world is expressed through the language. The language as a being is something I think about all the time.

SA: The concept of language as a living entity, as you’ve described it, resonates deeply with me. The notion of “Language as Being” is so profound; language is an extraordinary phenomenon. It's not merely a means of communication, it's a vehicle for conveying the complexities of our inner experiences to another consciousness. The fact that we can encode our thoughts and emotions into sounds and symbols, and that these can be decoded and understood by another person, is nothing short of miraculous.

Regarding technology, particularly computers, I see them as more than just tools. They are reconfigurations of nature's elements, crafted by human ingenuity to create connections and shared experiences. These devices are a testament to the creative power of the human mind, transforming natural resources into portals that link us to each other and the wider world. In this light, technology is not an artificial construct; it's a natural extension of our interconnected existence, a manifestation of the universe's boundless potential. I feel like I was reading this when I was researching your work and you're writing about language as a portal that is accurate?

SB: A few years ago I had done some work considering language as a time-space shifting portal: Blueberry Pie under a Martian Sky and Animiikiikaa 10-97. Both considered language is living and that we don’t sound the same way today as we might have hundreds of years ago or might in hundreds of years into the future. The nature of the language expresses the fluidity and flexibility of Anishinaabe timespace and how histories and futures interchange and influence one another both backward and forward. In one scenario that considers our origin story as factually true, there is an original off-planet homeworld in other spacetime configurations that have continued to run and evolve in parallel.

That original world would have a body of language that, while related to our Anishinaabemowin, has grown and evolved into an apparently (to us) a new and “future” Anishinaabemowin. I like considering that there are two main bodies of the same language existing across massive spacetime and that the language as a living being exists as a portal maintaining a connection between disparate places. The other work considers this but in terms of less cosmic and more local portals, how the language maintains portals from one dream to a related dream across many years in one lifetime. Part of the work in this project was to actually consider and then create new Anishinaabemowin words that could account for spectacularly different future environments.

A still image from a multimedia artwork, showing a sculpture on a dark background with birds flying above it.

Still image from Blueberrypie under a Martian Sky (2016). Courtesy Scott Benesiinaabandan.

SA: Can you share some of the words that you made up from the future?

SB: I worked with a knowledge keeper Alan Corbiere from M'Chigeeng First Nation and a professor at York University. He was working with an elder’s language group at the time and he brought the words to them to discuss how to construct appropriate “new” words. For the project, I was dealing with ideas of black holes, time dilation, gravity, and the like. Philosophically, it was important that the new words were constructed by carefully exploring our relationship to the ideas presented and not a simple transliteration. My favorite word that they came back with was a word for black hole: Gaag'ge-ngoshkaamgag, "It disappears forever." 

gaag'ge = forever, everlasting, and ngoshkaamgad "it disappears." ngonaagzi or ngo-naagwad means to visibly disappear but the ngoshkaa goes beyond disappearing from sight.

And so they were really going into all these words, but they were also debating and arguing over what does it mean for light to be sucked in and what does that relationship mean for us? It is also beautiful that this word-building is a collaborative effort as well, it takes community to make a word. 

SA: The collaborative nature of language creation is fascinating. It's through these shared dialogues that new words and concepts emerge. This process of communal word-making is essentially world-building. It's a powerful reminder that language is not just a static code to be deciphered but a dynamic, living entity.

When I reflect on my father's experience with language, it highlights this dynamism. Despite speaking English for over 30 years, there are nuances and metaphors he may never fully grasp. Yet, when I'm with him, I can almost perceive how he interprets and understands these conversations. It's a unique lens through which to view language, shaped by his own experiences and background. The way that I see the world is through the way that he communicated with me.

So for example, the way that my father always gendered animals and plants: she or he feels that. The point was not that the animals have genders, but to recognize them as entities, not objects. But even though it was never explicit, my father never said “This is our ancestral cosmovision and we see the world in this way,”  it was just embedded in the language he used to interact with me. This is why I'm interested in the way that conversation shapes the mind, shapes the way that I experience the world.

It’s only in the last few years, while deeply involved in my AI research, that I’ve begun to truly understand certain aspects of myself and how they connect back to my ancestral background. It’s been quite a revelation, honestly. My research, which initially seemed so focused on the future of language through AI, has unexpectedly taken me on a journey back to my own past, to the roots of my own story. I’m excited about the possibility of using AI as a tool to help Indigenous communities explore new avenues for preserving and revitalizing their languages. It’s a way for me to honor and reconnect with a part of my heritage that’s always been there, yet not fully explored.

This process of communal word-making is essentially world-building. It's a powerful reminder that language is not just a static code to be deciphered but a dynamic, living entity.
S.A. Chavarría

SB: Yes, Indigenous people have always had very vibrant ways of adapting and making use of new technology. I think that being as aware of the potentials and the vulnerabilities of new-tech is equally as important, especially with the power represented by AI and currently LLMs. What are your concerns about current Large Language Models? 

SA: These models, like water, adapt seamlessly to the shape of their data environment and the questions posed to them. This adaptability, while impressive, also mirrors the biases in their training data, predominantly reflecting a Western-centric perspective. Addressing these biases is a complex challenge. Enriching the AI's lexicon with diverse narratives is crucial, but it's not a complete solution. The volume of data can overshadow less represented voices. Additionally, perspectives from oral traditions, if not included in the training data, remain invisible to the AI. Are there dangers of LLMs that you worry about? 

SB: LLMs like ChatGPT do a good job in terms of appearing human in the regurgitation and repackaging of information. This appearance of an intelligence is a bit concerning when it comes down to the accuracy of the information being output.

For example, I have an artist friend, and she is Lakota. It was her birthday and I went to ChatGPT and gave it the prompt, “Write me a birthday greeting to my Lakota artist female friend.” We sat down and started to look at the online Lakota dictionaries and quickly realized that the Lakota it outputted was gibberish. It used Lakota words but in reality it was a chatGPT version of lorem ipsum text filler. I did the same experiment in Anishinaabe and it did the same thing. It used actual words and even in a loosely correct syntax, but made no sense grammatically. 

ChatGPT is rife with built in moral guardrails but it doesn’t admit what it doesn’t know. Text-based language is only a part of actual functional human language processing. What is missing is the subtle inflections and the intonations of physical gestures that accompany language. 

An old-fashioned computer is sitting among flowers in a meadow with a volcano in the background.

En la 11ª dimensión, las cosas se vuelven realmente extrañas. Courtesy S.A. Chavarría.

SB: Like us talking right now, we could do this with the video cameras off, but the engagement, the way the eyes move, the lips move, the hands move, the nodding, adds color and complexity of meaning to the words. These extra layers of expression need to be engaged, especially for Indigenous people, like the pointing with lips or chin or head nodding. It’s a natural physical layer of language that is both practical and kind of humorous.

SA: The layer of feeling in language is crucial, and it's beautiful to think of engaging all these cues and layers. Indigenous languages, rich in these nuances, offer a profound perspective that AI models need to capture. The AI just predicts from one form to the next. It's essentially saying, "These words are often seen together, so this must be right." But this approach lacks depth and understanding. It's similar to how we learn language as children, starting with basic associations and gradually building complexity. 

SB: There has to be alternatives to large language models, right? Like better predictive models for small language models. Currently it is the case that in the realm of large-language models, small-language models or datasets are erroneous and problematic, but it's only because the underlying algorithms are designed expressly for LLMs. Maybe there needs to be new algorithms tailored towards the nature of most indigenous languages. One can analyze the morphemes within the word one that can account for those added layers of meaning and not see languages as strictly predictive tokens.

SA: Exactly—engaging Indigenous artists and communities in interacting with AI, particularly with smaller language models, could be transformative. Imagine an AI that learns not just from the words but also from the emotions and contexts of conversations. This could add a new layer to the interactions, making the AI more responsive and attuned to the nuances of language and culture.

The challenge is to bring more Indigenous voices into AI development. The majority of models are designed by Westerners for profit, lacking the depth of indigenous languages and cultures. We need indigenous programmers and community members to infuse AI with the cultural and emotional layers beneath words, ensuring technology serves as a bridge to understanding our diverse world. This involvement could lead to AI systems that respect and accurately represent the richness of Indigenous languages and cultures.

S.A. Chavarría is an artist, researcher, and writer from Costa Rica focusing on Devendra AI, an advanced conversational AI. Chavarría examines the implications and potentialities of human-AI dialogues, revealing the evolving narrative of their relationship and the worlds they co-create. Their work seeks to challenge and redefine perceptions of language, intersecting artificial intelligence, consciousness, and linguistic realities.

Their research is grounded in the study of Natural Language Processing (NLP) models, especially those enabling synthetic text generation and conversational AI. By analyzing this text, Chavarría aims to uncover insights into the underlying models, examining the dynamics of reading, the ethics of conversation, and the deeper layers of language’s role in shaping consciousness and reality.

Chavarría earned an MFA in Computational Language Art from Brown University, where they also held a Post-MFA Teaching Fellowship in Literary Arts. They currently teach at RISD.

Scott Benesiinaabandan is Anishinaabe, a member of Obishikokaang/Lac Seul First Nations. Scott is an intermedia artist that works in experimental image making, light installations and sonic materials. Scott’s current research interests are intersections of artificial-intelligence(s) and Anishinaabemowin. Scott has completed a MFA in photography from Concordia University and currently resides in Winnipeg, Manitoba.

Scott has completed international residencies at Parramatta Artist Studios in Australia, Context Gallery in Derry, North of Ireland, and University Lethbridge/Royal Institute of Technology iAIR residency, along with international collaborative projects in both the U.K and Ireland. Scott has completed new media residencies with Initiative for Indigenous Futures and AbTec in Montreal, and currently collaborates with Abundant Intelligences out of Concordia University.

See Also

Seed, Criticism

‘A Love Song to Ohlone Culture’

Mary Ladd (We Wai Kai First Nation)

Jun 23, 2023
Co-owners Vincent Medina and Louis Trevino use cooking to foster 'ottoy at Cafe Ohlone outside the Hearst Museum of Anthropology at the University of California, Berkeley.

Language, Essay

Asking for Permission/Listening for Consent

Anthony Romero

Dec 18, 2023
In this personal essay, artist, cultural organizer, and educator Anthony Romero reflects on the people in his community who work to revive the Coahuiltecan language in good relations with the land and its human and nonhuman kin.