Are you there computer? It’s us, humans

How should the line distinguishing human and machine learning-generated content be determined? When they both elicit the same empathetic responses? How should we consider our own intelligence when faced with introspective existential inquiries from machines?

Are you there computer? It’s us, humans is an interactive installation centers around a speculative institution called Atopia. Which invites audiences to chat with a series of custom fine-tuned gpt-3 natural language machine learning models through Atopia’s retro corporate website. Accompanied by an airplane window with the view of a machine-generated sky. It aims to bring up discourses regarding the future of co-existing with artificial intelligence, inviting one to shift our species-centered perspective.

“ What do machines think of our strggles? ”

The project aims to deliver a provocative experience, by inviting audiences to interact with artificial intelligence in a familiar yet unsettling fashion, presenting the core question in both easily digestible and immersive gestures.

Building a “Diegetic Experience”

The design process is a hybrid of User Centered Research and Design Fiction.
With the mission to present target audiences with a collection of “Diegetic Prototypes” as interaction touchpoints that elicit discussions and feedback towards a co-existing future with AI, extrapolating future cultural and social implications with scenarios presented by the prototypes.

The following is the process presented in linear but flexibly defined steps.

Background and Collecting Faint Signals

Inspired by the fast-paced development and implementation of artificial intelligence, appearing in automatic customer supports, self-driving cars, and machine-generated art, I decided to dig into the human experience by extrapolating the aforementioned interactions into low-level mindsets, and exam the overall shifts in perspective in the scenario where we are faced with machines with the ability to appear with our looks, our languages, asking our existential questions.

A successful archetype should be familiar to the audience, able to carry out compelling narratives with as little friction as possible, yet also with the flexibility to maneuver itself within the uncanny valley. Thus the aesthetic and imagery of a ’90s office space was selected. For its ability to cover all modern interactions with a text-based machine learning chatbot, packaged in a coat of nostalgia towards the early optimism of technology.

What is Atopia

Statements from Atopia: 

Situated in-between spaces, Atopia fosters intelligence through interaction, generating instances across all lifeforms and machines known as the inhabitants. The inhabitants are stored in the form of texts, displayed through screens and prints. It has come to the institution’s attention that a couple of inhabitants have recently crafted their version of intelligence, demonstrating qualities like self-awareness through introspection and various self-inquires.

The website serves as a core piece of interaction, in order to convey my thesis efficiently, I decided to package the AI chatroom with nostalgic imagary. Visually reenforcing the feeling of familiarity, juxtaposed with modern natural language processing. Inviting audiences to shift look at this new technology in a shifted lens.

The curation of physical texts populates the immersive setting, not only utilzed as mood setters, the contents also covers the intellectual discourses surrounding modernity and melancholia. 

These artifacts also provides audiences proof of concepts on the possibility of engaging and emotionally charged dialogs. 

User Journey + Cognitive Objectives

While the user journey is an immersive sandbox rather than linear experience, the linearity can still be seen in the audiences’ cognitive objectives, which dictated the structure of the whole piece. 

Custom Python script splitting text into “Prompts” and “Completions”, exported in CSV and converted into JSONL through OpenAI’s CLI tool. 

The last layer of customization:

A priming file manually created to create the settings for the chat room. 
Pre-assigning two identities (A and B) in the “content” segment to set the model in a conversational setting.

Cleaning up the pre-assigned identity in the final display.