Artificial intelligence research has shown dramatic improvements in performance on natural language processing tasks such as machine translation, information retrieval, and question answering by feeding large corpora of natural language into machine learning systems. However, recent work on adversarial examples has shown that these systems may only be exploiting surface-level patterns in training datasets to get high scores on the tests instead of actually understanding language. This talk will discuss a particular thread of work toward building computer systems that understand natural language in depth and more like people do. The core method attempts to transform language inputs into language-free structures that are complex combinations of conceptual primitives, forming a substrate for human-like understanding, non-linguistic cognition, and reasoning processes that are not stated explicitly in language. Given time, he will also discuss using crowdsourcing to build a knowledgebase of these structures, the “inverse” problem of generating natural language output from them, and rich connections to image schemas, which form a parallel thread of work in cognitive linguistics.

Jamie Macbeth is an assistant professor in the Department of Computer Science at Smith College, and an MIT MLK scholar at CSAIL. Previously he was an assistant Professor of computer science and Elizabeth DeCamp McInerny Professor of Health Sciences in the School of Engineering at Fairfield University. In 2018 he was named Graduate Teacher of the Year at Fairfield University by Alpha Sigma Nu, the National Jesuit Honor Society. He received his Ph.D. in Computer Science from UCLA, and is a member of the ACM, AAAI, IEEE, and HFES.

All are welcome to join us for this educational luncheon. 

Please register here: https://www.eventbrite.com/e/mlk-visiting-scholar-luncheon-with-jamie-macbeth-tickets-83997283135

If you are able to attend, please reply to Shauna Bush-Fenty (sfenty@mit.edu) with any dietary restrictions, accessibility considerations or other needs.

​Institute Community and Equity Office (ICEO)