Ardavan Bidgoli to Present Doctoral Thesis Proposal Wed 18 Dec at 2pm

300x300 logo.jpg

Ardavan Bidgoli will present the proposal for the thesis, “Situated Toolmaking for Creative Computing: A Framework for Context-aware Machine Learning in Design, Art, and Making,” as a candidate of the PhD in Computational Design (PhD-CD) on Wednesday 18 December at 2:00pm.

Title: “Situated Toolmaking for Creative Computing: A Framework for Context-aware Machine Learning in Design, Art, and Making”
By Ardavan Bidgoli, PhD-CD Candidate

Date: Wednesday, 18 December 2019
Time: 2:00-4:00pm
Location: MMCH 121

PhD Advisory Committee

  • Dr. Daniel Cardoso Llach (Chair)
    Associate Professor, School of Architecture, Carnegie Mellon University

  • Dr. Eunsu Kang
    Visiting Professor of Art and Machine Learning, Carnegie Mellon University

  • Prof. Golan Levin
    Associate Professor of Art, School of Art, Carnegie Mellon University

  • Dr. Barnabás Póczos
    Associate Professor in Machine Learning, Machine Learning Department, Carnegie Mellon University


Abstract

Recent advancements in artificial intelligence and its sub-branch machine learning (ML) have led a growing number of artists, designers, and architects to explore these techniques’ affordances in developing tools that support their creative practices. However, the distinct separation between expert users and ML experts in the process of toolmaking results in abstract and decontextualized tools that are incapable of serving in real environments. This research proposes a paradigm shift in the process of ML-based creative computing toolmaking that accommodates expert users and contextual data as integral elements of the toolmaking process. It grants expert users more involvement and control over the process, mitigating the adverse effects of the decontextualization. The framework utilizes human-centered ML techniques such as interactive ML, Learning from Demonstration, and generative models to improve expert users’ ability to build their own creative computing tools and introduce contextual data into it without engaging with the complexities of the backend ML systems.

View the proposal document here.