Accessible only to conference ticket holders.
Log in Create account
For 90 days after a conference, only paid ticket holders can watch conference videos. After that, all Gold members have access.
If you have purchased recording access and cannot see the video, please contact support.
Moving AI offscreen: Exploring failures, constraints, and recovery in physical game design
This video is featured in the AI Conference playlist.
Summary
This case study explores what happens when generative AI moves off the screen and into a physical, walk-up-and-play game. Built by a team of four non-engineers, the project used AI to design and prototype everything from game logic to custom 3D-printed and laser-cut controllers. While AI accelerated early exploration, it repeatedly failed in physical space, where sensor noise, variability, and real players exposed its limits. The case study focuses on how designing for failure, constraints, and recovery ultimately mattered more than making the AI smarter, and what this reveals about trust, responsibility, and design judgment when AI becomes part of a real-world interactive system.
Key Insights
-
•
AI compresses years of training and enables access to unfamiliar design domains rather than replacing expertise.
-
•
Providing AI with rich contextual information via voice input improves output relevance significantly compared to short typed prompts.
-
•
Validating concepts cheaply with basic physical prototypes before investing in hardware avoids wasting resources.
-
•
AI can generate usable 3D design files and even scripting code to create printable models, accelerating prototyping.
-
•
Iterative back-and-forth with AI to refine outputs is a critical skill, not a single-prompt solution.
-
•
AI-generated solutions may work once but can fail in variable real-world conditions, necessitating user testing.
-
•
Physical hardware design forces more rigorous consideration of edge cases compared to software alone.
-
•
Tools like Adafruit's Circuit Playground Express enable physical inputs to simulate keyboard presses in software easily.
-
•
Scaling 3D prints requires adjusting slicer settings manually, a complexity AI cannot yet fully solve.
-
•
AI opens the door for creators to realize ideas they thought were out of reach, even without formal expertise.
Notable Quotes
"AI is not going to make you an expert but it compresses access to previously inaccessible domains."
"Don’t type with AI, talk to it like a brilliant colleague who knows nothing about your project yet—give it everything."
"Working with AI is a back and forth negotiation, not a vending machine where one prompt gets you the answer."
"The lesson isn’t that AI was wrong, but that you still have to test it against your users."
"I am not a hardware engineer, but I made something requiring those skills because AI gave me access to the doorway."
"In physical design, edge cases are more visible and testing immediately shows you problems you might have missed in software."
"AI can generate 3D models and geometry files, which you can directly import and print after careful checking."
"When scaling up 3D prints, slicer settings have to be adjusted—something AI didn’t help me with."
"A feature that only works some of the time is a feature that fails publicly in a walk-up installation."
"You don’t have to build a game; maybe you want to build a toy for your cat or a lamp—it just has to be yours."
Or choose a question:
More Videos
"Apple’s strength is the ease of learning, well-documented support, and consistent updates."
Sam ProulxUnderstanding Screen Readers on Mobile: How And Why to Learn from Native Users
March 27, 2023
"We are all multitudes of identities and experiences, far beyond what any screen can show."
Jayne Engle Tanya Chung-Tiam-FookCivic Design for the Next Seven Generations—A Discussion on Sacred Civics
August 25, 2022
"If you don’t trust it, then there’s nothing here."
Daniel J. RosenbergDesigning with and for Artificial Intelligence
August 11, 2022
"The old way is over. It's over."
Erika FlowersThe Handoff is Dead: Design-Led Engineering with AI Agents
March 4, 2026
"Our foundational interview, observation, and sense-making skills are evergreen and indispensable in the AI world."
Liwei DaiThe Heart and Brain of the AI Research
March 31, 2020
"Users left data trails—data exhaust—that can actually inform and empower them if designed well."
Sam LadnerData Exhaust and Personal Data: Learning from Consumer Products to Enhance Enterprise UX
June 8, 2016
"Find your senior executive ally who understands design and go talk to them first — it works."
Amy MarquezINVEST: Discussion
June 15, 2018
"If we improve areas of friction on behalf of the design department, people’s ability to achieve goals will increase."
Jacqui FreyFlow and Superfluidity for Design Orgs
November 7, 2018
"That night at 11:59 PM the entire nation of New Zealand would move into what was called a level 4 lockdown."
Brendan JarvisFraming Tomorrow by Questioning Today
June 8, 2022
