Ben Jones - Enabling AI in Computer Aided Design Through Representations

Modern computer aided design (CAD) tools evolved to support design
workflows that refine general ideas into concrete implementations.
Crucial to this is their choice of underlying representations;
parametric and procedural models that are realized as concrete geometry
for evaluation. In the first part of the talk, I will showcase machine
learning systems that work in concert with these representations to
enable AI assistance in existing design workflows. Design is, however,
an iterative process, and currently computational tools leave the
majority of the design cycle in the mind of the designer. I will show
how existing tools enabled AI assistance across a single iteration of the design cycle and present an argument and vision for why procedural generative models are a great fit for adding computational assistance across the full process of design. I will end by discussing nascent work attempting to realize this vision across a variety of design domains.

Bio:
Ben Jones is a postdoc in the MIT Computational Design and Fabrication Group advised by Wojchiech Matusik. He earned his PhD in Computer Science and Engineering at the University of Washington advised by Adriana Schulz, and Bachelors of Science degrees in Physics and Mathematics and Computer Science at Harvey Mudd College. His research interests are broad, and focus on applying computation to support invention and discovery. Currently Ben explores applying machine learning to design and proof synthesis, and in the past has worked on parametric CAD systems, automated knitting, distributed systems, computational microscopy, and space-based solar power.