Faculty Recruiting Support CICS

Towards Literate Artificial Intelligence

09 Apr
Tuesday, 04/09/2019 4:00pm to 5:00pm
Computer Science Building, Room 151
Special Event
Speaker: Mrinmaya Sachan

Abstract: Over the past decade, the field of artificial intelligence (AI) has seen striking developments. Yet, today's AI systems sorely lack the essence of human intelligence i.e.  our ability to (a) understand language and grasp its meaning, (b) assimilate common-sense background knowledge of the world, and (c) draw inferences and perform reasoning. Before we even begin to build AI systems that possess the aforementioned human abilities, we must ask an even more fundamental question: How would we even evaluate an AI system on the aforementioned abilities? In this talk, I will argue that we can evaluate our AI systems in the same way as we evaluate our children - by giving them standardized tests. Standardized tests are regularly administered to students to evaluate the knowledge and the skills gained by them as they progress though the formal education system. Thus, it is a natural proposition to use these tests to measure the intelligence of our AI systems as well. Then, I will describe Parsing to Programs (P2P), a framework that combines ideas from semantic parsing and probabilistic programming for situated question answering. We used the P2P framework to build two systems that can solve pre-university level Euclidean geometry and Newtonian physics examinations. P2P achieves a performance at least as well as the average student on questions from textbooks, geometry questions from previous SAT exams, and mechanics questions from Advanced Placement (AP) exams. I will conclude by describing implications of this research and some ideas for future work.

Bio: I am a PhD candidate in the Machine Learning Department, School of Computer Science at Carnegie Mellon University. Eric Xing is my fantastic adviser. My research is in the interface of machine learning, natural language processing, knowledge discovery and reasoning (inference). Specifically, I focus on machine reading -- on building background knowledge of the world by extracting knowledge from text and incorporating the extracted background (domain) knowledge to solve problems that require complex reasoning. During my doctoral studies, I have focused on building automated solvers for standardized tests such as the SAT and the advanced placement (AP) tests. These tests are regularly administered to students to objectively measure their knowledge and skills. Thus, it is natural to see these tests as benchmarks (and hopefully as drivers of progress) for AI. I am also interested in using these solvers as assistive technologies that can teach students. You can view my publications to get a more definitive idea of my work.

I spent my undergraduate years at Indian Institute of Technology (IIT), Kanpur, where I graduated with a B.Tech. in Computer Science and Engineering. I worked with the Information Management group at IBM Research India from 2010-12. I spent a fantastic summer in 2014 at the Machine Learning Group at Microsoft Research, Redmond where I worked on Machine Comprehension with Matthew Richardson. This work was selected as one of the outstanding papers at the ACL 2015 conference. I also spent a lovely summer in 2016 at the Allen Institute of AI where I worked with the Euclid team. I am fortunate to be supported by the CMLH Fellowship for the year 2017-18. I was supported by the IBM Fellowship in 2016-17 and the Siebel Scholarship in 2013-14. I was also a finalist for the Facebook Fellowship in 2014-15. I regularly review for top ML and NLP conferences.

Besides my research, I am passionate about sports (particularly football - read soccer ;-) , tennis and squash), socio-tech startups and community service. I like to listen to music (contemporary, instrumental, indian classical, sufi, ghazals, electro, etc.) and travel around the world.

A reception for attendees will be held in CS 150 at 3:30 p.m.

Faculty Host
: