Invited Talk by Sridhar Mahadevan
Talk Title: Imagination in Humans and Machines
One of the hallmarks of human intelligence is the capacity to imagine — defined as the ability to transport oneself in time, place, and circumstances. Humans exhibit a remarkable capacity to imagine from an early age — children are able to invent games, think of impossible objects, and other imaginative feats. In this talk, I will describe how human-inspired imagination can be transported to machines, and provide the basis for the next generation of advances in AI and machine learning. Specifically, I will summarize ongoing work on causal imagination, which goes beyond correlative data mining to include causal interventions and counterfactual reasoning. I will also discuss recent work on visual imagination, using generative adversarial networks (GANs) and their variants. I will finally discuss work on linguistic imagination, including analogies and metaphors, and describe some abstract mathematical models of how to construct imagination spaces.
Sridhar Mahadevan has recently become Director of the Data Science Lab at Adobe Research in San Jose, CA. Adobe’s Digital Experience Cloud processes hundreds of petabytes of web analytics data, and provides a valuable platform for exploring large scale data science. He is also a Visiting Professor at the Stanford University School of Engineering, and a Professor at the University of Massachusetts, Amherst. He was elected Fellow of AAAI in 2014 for significant contributions to machine learning. He has published over 150 papers in many areas of AI and machine learning. His AAAI 2018 paper on imagination machines received a Blue Sky Best Paper award from the Computing Community Consortium. He is also giving a tutorial on imagination machines at IJCAI-ECAI 2018.