About the Talk "Towards Generalizable Autonomy: Duality of Discovery and Bias": Generalization in embodied intelligence, such as in robotics, requires interactive learning across families of tasks and is essential for discovering efficient representation and inference mechanisms. Concurrent systems need a lot of hand-holding to even learn a single cognitive concept or a dexterous skill, say “open a door”, let alone generalizing to new windows and cupboards! This is far from our vision of everyday robots! would require a broader concept of generalization and continual update of representations.
My research vision is to build the Algorithmic Foundations for Generalizable Autonomy, which enables robots to acquire skills in cognitive planning & dexterous interaction, and, in turn, seamlessly interact & collaborate with humans. This study of the science of embodied AI opens three key questions: (a) Representational biases & Causal inference for interactive decision making, (b) Perceptual representations learned by and for interaction, (c) Systems and abstractions for scalable learning.
Reinforcement Learning as a paradigm provides a powerful mechanism to use large scales of data with minimal intervention, replacing interaction for supervision. However, many of the concurrent methods are both compute and data-intensive, both of which can be bottlenecks in applied settings. In this talk, we will describe (a) Model-Based RL: how to use data with structured models (CAVIN, SHAC) and objectives (VaGRAM) (b) Offline RL (IRIS) and data augmentation (CoDA, S4RL).

About the Speaker: Animesh Garg is a CIFAR Chair Assistant Professor of Computer Science and Mechanical Engineering (courtesy) at the University of Toronto, a Faculty Member at the Vector Institute and UofT Robotics Institute, where he leads the Toronto People, AI, and Robotics (PAIR) research group. Animesh is also a Senior Researcher at Nvidia Research. Animesh earned a Ph.D. from UC Berkeley and was a postdoc at the Stanford AI Lab. His work aims to build Generalizable Autonomy which involves a confluence of representations and algorithms for reinforcement learning, control, and perception. In particular, he currently studies three aspects: learning structured inductive biases in Sequential decision making, using data-driven causal discovery, and transfer to real robots: all in the purview of embodied systems. His work has received AAAI New Faculty Highlight and Best Paper Recognitions at top tier venues in Machine Learning and Robotics such as ICRA, IROS, RSS, Hamlyn Symposium, Workshops at NeurIPS, ICML, and has been widely covered in the press New York Times, Nature, Wired, IEEE Spectrum.