The ethnography of prompting

Formulating is more than asking

Language and Perception

As a species, we have always used our sense to produce language in order to explore and capture various aspects of the world, from weather, rivers, plants, and animals to human interactions. The act of producing meaning not only connects us with the external world but also allows us to delve into the collective knowledge and perspectives passed down through generations.

Symbolic Technologies

Symbolic registration of our perceptions dates back to ancient times, as evidenced by archaeological finds like the Lebombo and Ishango Bones, the oldest mathematical objects discovered. Since then, our ability to store and interpret symbols has become deeply ingrained in our species. We extend our cognition through the use of materials to create objects, mark surfaces, and adorn ourselves. The internet has evolved into a synthetic super organ, transcending its physical presence.

The Nature of Language Models: As we increasingly rely on digital tools, our external capabilities expand, while internally we seek the same pleasure we derive from our senses. We marvel at the seemingly instantaneous thought machines display when responding to our inquiries. The supernatural symbolic power bestowed upon these technologies surpasses our comprehension, thanks in part to media portrayals.

Ethical and Societal Implications

In our changing cultural landscape, information has replaced tradition as a driving force. While language models contribute to language generation, it is important to acknowledge that they are ultimately designed and developed by humans. The collaborative nature of human-machine interaction in language production and its cultural aspects should be emphasized to gain a more accurate understanding. The ethical concerns arise from the biased algorithms, low-wage labor, and challenges associated with explainability.

Language and Communication

Language models, like LLMs, are tools designed and developed by humans to aid in language generation. Understanding their cultural outputs requires addressing the algorithmic biases, the role of human labor in data collection and model training, and the challenges associated with interpreting machine-generated language. We should move beyond assumptions of “lack of subjective experience” and explore our fascination with these systems, considering the societal disparities they may perpetuate.

The weaving of public consciousness

In our digital era, the concept of “organic-like” products has emerged, from genetically modified organisms to artificial meat. Image, sound, speech, and thought are now digitized, existing but intangible. Language production, beyond mere prompting, has its limits. Machines cannot surpass what is prompted, and yet, there is a sense of “instantaneous knowledge” associated with them. Our own language rules, akin to Michelangelo’s marble statue, contribute to this enchantment.

By delving into the concept of “organic-like” in the context of language generated by non-organic sources, we gain a clearer understanding of the phenomena observed in language models’ hallucinatory “behavior.” Rather than focusing solely on the limitations and differences between machine-generated language and human language, we can embrace the enchantment they evoke, recognizing the underlying human-written code and its social disparities.