Abstract: Deep generative models have made significant progress in a variety of real-world applications. However, these models are often interpreted as black-boxes without the ability to understand what knowledge they learn in their latent representations. Before we can widely deploy deep generative models in high-stakes tasks such as self-driving cars or medical diagnosis, we need to understand the decision-making process of these models. This is an important research question with many practical implications, allowing us to 1) explain what knowledge generative models learn and what kind of controls they provide (ICCV’21, WACV’22), 2) facilitate new forms of creativity and human-AI collaboration (WACV’23, NeurIPS WS’22, CVPR WS’22, ACM CC’22), 3) enable more transparent and fair models by identifying and fixing potential biases (ECCV’22).

In this talk, I will present our recent work from CATLAB on interpreting and manipulating the latent space of GANs (ICCV'21, WACV'22). I will also talk about generative models in creative content generation and discuss human-AI collaborated designs as part of the How to Generate (Almost) Anything project (https://howtogeneratealmostanything.com). Finally, I will briefly discuss our recent efforts to address issues related to fairness and ethics in generative models, and conclude my talk with future work.

Bio: Pinar Yanardag Delul (https://pinguar.org) received her PhD in Computer Science from Purdue and completed her postdoctoral research at MIT Media Lab. She is currently a principal investigator at Bogazici University, Istanbul, Turkey, where she leads CATLAB (https://catlab-team.github.io). Her research interests are in the area of generative adversarial networks (GANs) and developing deep learning approaches to enable creativity and human-AI collaboration. Her research has been published at conferences such as NeurIPS, ICCV, and KDD, and featured in mainstream media (e.g., The Washington Post, BBC, CNN) and magazines (e.g., Motherboard, Rolling Stone). She is also CEO of AI Fiction, a creative design studio focused on the entertainment industry, for which she received a Creative Director nomination at the 72nd Primetime Emmy Awards for her work on the Westworld TV series. During her studies, she was supported by Fulbright Fellowship, and Google Anita Borg Scholarship.

RSVP for Zoom information: https://pitt.co1.qualtrics.com/jfe/form/SV_bEnPpfqNmncp8RE