Adarsh Pyarelal, PhD

ADARSH AND WIN'S TALK [WATCH ON YOUTUBE]:

As anyone who's been frustrated by Siri or Alexa can readily verify, computers just don't get us. While artificially intelligent (AI) agents are getting quite good at understanding explicit instructions, they still struggle to understand implicit information conveyed through prosody, facial expressions, informal/imprecise language, etc. This difficulty presents a major obstacle to the development of AI 'theory of mind', i.e., the ability to infer the beliefs, desires, and intentions of humans. In this talk, Adarsh will give an overview of ToMCAT (Theory of Mind-based Cognitive Architecture for Teams) – a 4-year project aimed at developing AI agents with social intelligence and testing them in a Minecraft-based virtual testbed, and Win will speak to the future applications of this work.

ABOUT ADARSH:
I am a research scientist in the School of Information at the University of Arizona. My research interests include the development of socially-aware AI teammates, and the automated assembly of executable causal models of complex systems from text and software. Broadly speaking, I am interested in machine learning and artificial intelligence. In particular, I am interested in Bayesian generative approaches to inference problems. I am currently the principal investigator on the ToMCAT project, funded by a DARPA grant as part of the Artificial Social Intelligence for Successful Teams (ASIST) program.