AI Talks

with Matthew Guzdial

Date: 28 May 2020
Time: 19:00 - 20:00 GMT

About the episode

About the talk

What does it mean for someone or something to be creative? A creative person is someone who can generalize from a few examples, produce highly novel solutions (compared to existing work), and explain their creative process. In machine learning, we often seek algorithms and models that share these properties. Ideally, we want models that generalize well from a few samples. When using machine learning to produce content, as in the case of image generation, we want models that can produce solutions dissimilar from their training data. We want models that are easily explainable, especially for users without technical knowledge. In this talk I will present my work investigating how to adapt machine learning to solve problems we’d typically expect to require human creativity (also called computational creativity). Specifically, I will focus on the challenge of automated video game design. Games challenge modern machine learning as they are interactive pieces of software that vary wildly from one another, without a single common representation. However, computational creativity is not just for games. I discuss my work on computational models of cognitive creativity to improve transfer and generalization in image classification. Finally, this capacity for automated creativity isn’t useful unless it directly benefits people, however it is still unclear how best to design these interactions. I will discuss ongoing research into this area of creative machine learning and human interaction. As computational creativity systems are growing in sophistication, I look at the question of how we can pair human creators and algorithms to support human creativity.

About Matthew

Matthew Guzdial is an Assistant Professor in the Computing Science department of the University of Alberta. His research focuses on the intersection of machine learning, creativity, and human-centered computing. This includes investigating machine learning-based automated game generation, human-AI design collaboration, improving machine learning transfer and generalization with computational creativity, and user experience modeling. He is a recipient of a Unity Graduate Fellowship and two best conference paper awards. His work has been featured in the BBC, WIRED, Popular Science, and Time.