This is a collaborative design research project for CMU's School of Design Capstone

Motivated by the infiltration of Artificial Intelligencce into our daily digital interactions, my team and I designed an educational arcade game system to teach about Deepfakes and the GANs algorithm.

collaboration w/	Ethan Ye, Sherry Wu, and Jake Zimmer
tools	hardware prototyping raspberry pi custom CAD and fabrication javascript python
skills	project management, human-centered research, physical computing

This project focused on deepfakes and how it influences our understanding of identity through the lens of technology. We devised multiple research exercises to assess subject's ability to differentiate real from fake faces

WHAT IS?

From our research, we learned users couldn't recognize AI generated faces as well as they thought they could, we designed this system of arcarde games to educate users. The system-- composed of two parts-- represents the two phases in the GANs algorithm: generator and discriminator, only now represented in the form of two arcade fames. AI Invaders represents the generator aspect of the GANs algorithm, inviting users to manipulate and witness, right before their eyes, faces being generated from those of the school of design senior class. When content with the face they had created, the newly generated face gets sent to the second part of our system-- AI Defenders. Here, we invited users to differentiate which faces were real or fake.

The games allow users to tangibly experience being both the creator and reviewer of AI-generated content. This experience was designed to bring about more AI literacy and spark conversation about the rapid development of this technology

HOW?

Each game interface was built in Javascript, featuring 3D printed hardware and MDF housing. The design of the housing was conceived to lower the barrier of entry to understanding AI and present the information in a tangible and fun way. Since AI Invaders required quite a bit of computational power in order to see the faces morphe smoothly, we hid a GPU inside the machine and integrated our software via a Teensy 3.2 receiving input from our physical sliders and buttons. The entire AI Defenders game was built in javascript and linked to a google drive where AI Invaders was sending the newly generated faces. Users could differentiate real from fake via 3D printed hardware running off a raspberry pi.

DOCUMENTATION LINKS

* In-depth Process Documentation
* Final Presentation
* Preliminary Research
* Gallery of Generated Faces from AI Invaders

MY ROLE

Beginning in my sketchbook, I worked adaptively from concept to completion and led the fabrication of all aspects physical in this project. I designed, CAD'd, CNC'd, 3D printed, sanded, and painted the housing and physical interface. In addition, I led the integration of software with hardware via raspberry pi and Teensy with Jake and the development of a cohesive visual style to consistently feature in all aspects of the game system.