Can you trust an LLM?

Day 2 | 11:00 – 12:00 | Main Hall

Photo. Portrait of Milica Gašić.

Prof. Dr. Milica Gašić

Heinrich Heine University Düsseldorf

Abstract

Dialogue systems facilitate interaction between a human and a computer in natural language. Building dialogue systems a couple of decades ago was a niche area, but today’s largest dialogue systems—large language models—are becoming a part of our daily lives, impacting education, science, medicine, and more. The promise and perils of LLMs lie in their data. While there has been a paradigm shift in natural language processing from building small expert models trained on labeled, homogeneous datasets to large language models trained on largely unlabelled and heterogeneous data, which are capable of solving a variety of tasks, issues like overconfidence and bias persist. In this project, we aim to develop methods for estimating uncertainty in LLMs as well as harness the insufficiently explored benefits of reinforcement learning, as we believe these are critical steps toward building trustworthy interactive systems. 

Prof. Dr. Milica Gašić

Milica Gašić is a Professor of the Dialog Systems and Machine Learning Group at the Heinrich Heine University Düsseldorf. Her research focuses on fundamental questions of human-computer dialogue modelling and lie in the intersection of Natural Language Processing and Machine Learning. Prior to her current position she was a Lecturer in Spoken Dialogue Systems at the Department of Engineering, University of Cambridge where she was leading the Dialogue Systems Group. Previously, she was a Research Associate and a Senior Research Associate in the same group and a Research Fellow at Murray Edwards College. She completed her PhD under the supervision of Professor Steve Young and the topic of her thesis was Statistical Dialogue Modelling for which she received an EPSRC PhD Plus Award. She holds and MPhil degree in Computer Speech, Text and Internet Technology from the University of Cambridge and Diploma (BSc. equivalent) in Mathematics and Computer Science from the University of Belgrade. She is a member of ACL, a member of ELLIS and a senior member of IEEE as well as a member of the International Scientific Advisory Board of DFKI. 

Best paper awards: CSL (2010), Interspeech (2010), SLT (2010), Sigdial (2013), Sigdial (2015), EMNLP (2015), ACL (2016), EMNLP (2018), SIGDIAL (2020) and *SEM (2020).