AI researcher Leora Morgenstern talks about some hard problems in AI including pronoun disambiguation, knowledge representation, and creating AI solutions in the enterprise with incomplete data. She started her career in academia, but then realized that the problems in the real world had noisy data that provided additional challenges for AI that needed to be solved. Morgenstern was the lead author of an article called “Planning, Executing, and Evaluating the Winograd Schema Challenge,” which appeared in the Beyond the Turing Test issue of AI Magazine in the Spring of 2016. The Winograd Schema aims to create pronoun disambiguation problems that are easy for humans with common sense to solve, but that are hard for AI. We talk about her work with reading table information in microbiology journals, and her career of working on other knowledge representation issues that inspired her to help run the Winograd Schema challenge, which debuted at IJCAI 2016 and just had it’s second iteration during AAAI 2018.

This is a listener-supported podcast through the Voices of AI Patreon.
Theme music collaboratively composed by Melodrive‘s AI & creators.