Abstract: In this talk, I will present some of the Deep Reinforcement Learning (Deep RL) research that our group at Ubisoft La Forge has been working on. I’ll begin by giving an overview of Deep RL and highlighting some of its applications in video game development. I’ll then go into detail about some of our recent approaches including a deep dive into the problem of navigating complex environments. The talk will then conclude with a discussion of some first-hand complications/limitations of Deep RL that we’ve experienced.

Bio: Joshua Romoff is an R&D Scientist at Ubisoft La Forge, where he works on pushing the state of the art of Deep Reinforcement Learning for video game development. He completed his PhD at McGill university where he focused on methods for decomposing the Reinforcement Learning problem into simple components.

Abstract: Artificial intelligence has evolved to include convolutional neural networks or CNNs that allow for the extraction of a wide range of features from images using deep learning. This presentation explores the emergence and use of “neural algorithms of artistic style” or image style transfer and post estimation/detection to create diverse works of art.

Bio: Dr. Nettrice R. Gaskins is an African American digital artist, academic, cultural critic and advocate of STEAM fields. In her work she explores "techno-vernacular creativity" and Afrofuturism.

Dr. Gaskins teaches, writes, "fabs”, and makes art using algorithms and machine learning. She has taught multimedia, computational media, visual art, and even Advanced Placement Computer Science Principles with high school students who majored in the arts. She earned a BFA in Computer Graphics with Honors from Pratt Institute in 1992 and an MFA in Art and Technology from the School of the Art Institute of Chicago in 1994. She received a doctorate in Digital Media from Georgia Tech in 2014. Currently, Dr. Gaskins is a resident in the Autodesk Technology Centers Outsight Network. She is the assistant director of the Lesley STEAM Learning Lab at Lesley University. Her first full-length book, Techno-Vernacular Creativity and Innovation through The MIT Press will be available in August.

Gaskins served as Board President of the National Alliance for Media Arts and Culture (The Alliance) and was on the board of the Community Technology Centers Network (CTCNet). She is currently on the board of Artisan’s Asylum.

Abstract: At Rebellion we have redesigned our automated testing framework that goes further than most automated testing solutions; we developed a system that plays entire games from start to finish like a player would. In this talk we present this framework and explore how it combines an AI Planning system with classical techniques, like behaviours trees. We will discuss how we managed the complexity of the system and the helpful impact it has had on game development. We will showcase this system playing a variety of games from different genres, including: 3rd person shooters (Zombie Army 4) and a lair-management game (Evil Genius 2). We finish the presentation with an exploration of explainable planning.

Bio: Bram Ridder earned his PhD in Computer Science at King's College London where he researched AI Planning. During his Post Doctorate he worked on various robotic projects where he used AI Planning to create autonomous robots: the EU Project PANDORA created Autonomous Underwater Vehicles that can maintain and inspect subsea infrastructure; The EU Project SQUIRREL created autonomous robots that interacts with children, plays games with them, and uses an underlying cognitive model to determine how to interact with children. Four years ago he left academia and joined Rebellion to work as an AI Programmer on Zombie Army 4. In his free time he creates games using his own 3D engine and creates AI systems that play games autonomously.