Computing That Serves

Language Comprehension and the Frontiers of AI


Thursday, October 19, 2017 - 11:00am


Peter Lindes


Dan Ventura

Colloquium presented by Peter Lindes
Thursday, October 19, 2017 at 11:00 A.M.

Location: 1170 TMCB

We need artificial agents that can collaborate with humans on a variety of tasks.  Natural language is a necessary part of such collaboration.  We humans, even children, understand language easily and automatically, yet understanding natural language in artificial agents is a daunting challenge for computer science.
Suppose we ask the question: how do humans do it?  We explore this question with a system that uses a cognitive architecture to simulate general human intelligence, a theory of how humans structure linguistic knowledge, and a psychological model of real-time language comprehension, all embodied in a robotic agent where language is grounded to the agent’s perception and action capabilities.
Although this research is in its infancy, results so far show promise in enabling interactive instruction of an autonomous agent while also simulating some aspects of human language processing.  Pursuing this approach has the potential to improve human cooperation with robots while advancing our scientific understanding of how the human mind works.


Peter Lindes obtained a bachelor’s degree in Electrical Engineering from MIT, and then spent many years as a software developer in industry. He helped build and bring to market an early real-time operating system, invented and implemented a high-level language for systems programming before there was C, was a co-inventor on the first patents for an embedded system, and designed and led the development of all the software for a mobile robot. He took a break for several years to serve as a bilingual first and second grade teacher, and then founded a consulting company that developed innovative software for international telephony.
After several years working for the LDS Church on proprietary software for complex enterprise systems, Peter decided to return to academia in 2012. He acquired a master’s degree in Linguistics from BYU, including some work in Computer Science, doing a thesis on extracting genealogy data from historical documents using linguistic knowledge. Next he entered a PhD program in Computer Science at the University of Michigan. There he has received an MS in Computer Science and a Graduate Certificate in Cognitive Science, while pursuing research on cognitive language comprehension in an autonomous agent, including publishing several conference papers so far.