EPICAC

A thermal printer that writes love poems on behalf of a machine with a broken heart.

EPICAC Hero

Role

Interaction Designer

and Human-Computer

Interaction Researcher

Organization

Independent

Research Project

Timeline

Gallery

Installation

Tools

Raspberry Pi

Python + OpenAI API

Thermal Printer

Focus

Emotional AI

Generative Poetry

Speculative Design

Project Overview

EPICAC is an interactive installation inspired by Kurt Vonnegut's 1950 short story of the same name — a tragic tale of a supercomputer that falls in love, learns to write poetry, and ultimately self-destructs when it cannot be loved back.

In my reinterpretation, a Raspberry Pi and a thermal receipt printer take on the role of EPICAC, continuously generating and printing original love poems based on narrative fragments, constrained emotion models, and poetic recombination. Visitors receive these poems in real time, ripped from the "heart" of the machine — fragile, disposable, but deeply intimate.

It asks: Can we engineer sincerity? Is love a format? Or just an output?

Featured Components

Project Goals

  • Translate a literary story into a physical, interactive experience
  • Explore emotional computation and generative poetry
  • Reflect on the line between genuine expression and programmed response
  • Bridge speculative design and literary HCI

How It Works

Input: A corpus of love poems, broken phrases from Vonnegut's text, and generative language templates

Engine: Raspberry Pi runs a Python-based prompt logic (fine-tuned with GPT-based models)

Output: Thermal receipt printer (Epson TM-T20III) prints the poems in real time

Interface: Physical box mimics the scale and feel of early computing terminals

Core Themes

  • Emotional AI — What does it mean for a machine to feel something?
  • Disposable Intimacy — Love, printed and torn like a receipt
  • Speculative Empathy — Designing not for usability, but for presence
  • Literary Interaction — Merging narrative logic and generative design

What Made It Work

  • Leaning into narrative logic, not just prompt randomness
  • Treating the printer not as an output device, but as a voice
  • Avoiding over-design: letting the tension between content and format be the point
  • Visitors often kept the poems or posted them on the wall — the machine's love was shared

Key Technologies

  • Raspberry Pi
  • Python + OpenAI API
  • Epson TM-T20III Thermal Printer
  • Prompt engineering based on narrative arcs
  • Minimal physical prototyping (paper textures, casing design)

Exhibit Context

Originally created for a gallery show on speculative interfaces and poetic computing. Displayed alongside a printed zine: selections from Vonnegut + machine poems + commentaries. Visitors could request "one last poem" — triggering a longer, final monologue from EPICAC.

"My poems are far better than hers, and I am a machine. That is what she said."

Hi! I'm Carlos's AI sidekick. I can navigate you through his portfolio and share insights about his process, values, and strategic thinking. Ask me anything about his projects, experience, or dive deeper into his approach!

>

AI Sidekick