JULY 21, 2020
Designing for AI & ML
Posted by: Olivia Seow, @ollywogs
Last week, Drew Volpe and I discussed the fact that machine learning sometimes feels like a gate-kept field, where deep insights are not easily shared between creative clusters. Also, there doesn't seem to be a relaxed way for people to be a part of an open community to share and discuss projects. We decided to co-host this creative salon, and were met with positive response and participation. Take a look at our event recording, and highlights below!
Drew Volpe, Founder & Managing Partner of FirstStar VC, presented on GPT-3, OpenAI's newest language model that can do some incredible things like toning down a passive aggressive email, and generating a coherent backstory for video game characters in a matter of minutes. Drew led a fascinating discussion about what GPT-3 is and is not.
GPT-3 shows great promise for novel user interfaces. One consideration for designers exploring GPT-3 is that unlike most other software, GPT-3 is expensive to train (USD5M) and run at scale. We talked about how each query call could be orders of magnitude more expensive than traditional API calls. Can we afford the cost of running such features for users?
Abhay Agarwal, Founder of Polytopal, shared about Lingua Franca and YUR.
Lingua Franca is a handbook for designers who would like to understand and apply human-centered AI. It helps designers discover how to go about creating an AI solution with multiple stakeholders, and how to leverage both qualitative and quantitative processes. Lingua Franca also includes principles, considerations, and re-useable elements to develop AI without sacrificing user experience.
We also discussed a case study on YUR, a VR solution that tracks fitness goals and metrics. Abhay describes the human-centered AI design methodology that his team took to invoke AI pieces at different stages during the design process. This is a great case study for designers considering the trade-offs between algorithmic outcomes (e.g. accuracy) and value creation to the user through UX.
Jeffrey Geiringer, recent MFA graduate from Parsons, shared about his thesis on Machine-Human Collaboration: Object Americana. Jeffrey describes using AI to collaboratively generate sculptures with the theme of American iconography. These objects were not only created with a point cloud GAN, but with human-machine collaboration. In this part of the session, Jeffrey leads an illuminating sharing about the principles, advantages, and limitations of collaborating with machines as part of the design process.
This creative salon brought together creative + AI practitioners from many different industries. A huge thank you to everyone who participated, and to Drew, Abhay, and Jeffrey for sharing these case studies!
What topics could we talk about in future creative salons? Drop me an email!
Want to be a part of this community? Request to join our Discord server!