Public Lecture, Keynotes & Panel Discussion

The following speakers have graciously agreed to give public lecture, keynotes, and panel discussion at ALTA 2024.

Public Lecture - Monday 2nd December 5.30pm-6.30pm, Birch Building, ANU

Professor Eduard Hovy - University of Melbourne

LLMs: what they are and where they are heading

Register for the Public Lecture

Abstract

Generative AI has unleashed hype and concern. But it is surprising how few people understand how simple it is at heart, and how some of its shortcomings spring from its essential nature and will remain hard to overcome. In this talk I briefly describe the essential process and explore the three principal directions of GenLLM research: making them usable, useful, and understandable.

Bio

Professor Eduard Hovy is Executive Director, Melbourne Connect - a dynamic collaboration between leading organisations and interdisciplinary institutions aimed at leveraging research and emerging technologies to address global challenge - and a Professor in the School of Computing & Information Sciences, University of Melbourne.

Keynotes

Professor Jing Jiang - Australian National University

LLM Evaluation: Writing Styles, Role-playing, and Visual Comprehension

Abstract

Large language models (LLMs) have demonstrated exceptional abilities that extend beyond language understanding and generation. This underscores the need for a more comprehensive evaluation of LLMs that covers a broader spectrum of capabilities beyond traditional NLP tasks. In this talk, I will share some of our recent work on LLM evaluation, with a focus on LLMs’ writing styles and role-playing capabilities, and the abilities of large vision-language models to combine and interpret visual and linguistic signals in complex scenarios.

Bio

Jing Jiang is a Professor in the School of Computing at the Australian National University. Previously she was a Professor and Director of the AI & Data Science Cluster in the School of Computing and Information Systems at the Singapore Management University. Her research interests include natural language processing, text mining, and machine learning. She has received two test-of-time awards for her work on social media analysis, and she was named Singapore’s 100 Women in Tech in 2021. She holds a PhD degree in Computer Science from the University of Illinois Urbana-Champaign.

Professor Steven Bird - Charles Darwin University

Language Technology and the Metacrisis

Abstract

Despite their manifold benefits, language technologies are contributing to several unfolding crises. Small screens deliver mainstream content across the world and entice children of minoritised communities away from their ancestral languages. The data centres that power large language models depend on the mining of ever more rare earth metals from indigenous lands and emit ever more carbon. Malicious actors flood social media with fake news, provoking extremism, division, and war. Common to these crises is content, i.e. language content, increasingly generated and accessed using language technologies. These developments – the language crisis, the environmental crisis, and the meaning crisis – compound each other in what is being referred to as the metacrisis. How are we to respond, then, as a community of practice who is actively developing still more language technologies? I believe that a good first step is to bring our awareness to the matter and to rethink what we are doing. We must be suspicious of purely technological solutions which may only exacerbate problems that were created by our use of technology. Instead, I argue that we should approach the problem as social and cultural. I will share stories from a small and highly multilingual indigenous society who understands language not as sequence data but as social practice, and who understands language resources not as annotated text and speech but as stories and knowledge practices of language owners. I will explore ramifications for our work in the space of language technologies, and propose a relational approach to language technology that avoids extractive processes and centres speech communities.

Bio

Over the past three decades, Steven Bird has been working with minoritised people groups in Africa, Melanesia, Amazonia, and Australia, and exploring how people keep their oral languages and cultures strong. He has held academic appointments at Edinburgh, UPenn, Berkeley, and Melbourne. Steven established the ACL Anthology, the Open Language Archives Community and the Natural Language Toolkit, and is past president of the Association for Computational Linguistics. Since 2017 he has been research professor at Charles Darwin University, where he collaborates with Indigenous leaders and directs the Top End Language Lab, http://language-lab.cdu.edu.au. Steven pursues other language-related projects at http://aikuma.org.

Kyla Quinn - Australian Department of Defence

LLMs are great but …

Abstract

Knowledge workers are crying out for ways to industrialise the boring parts of their jobs, company executives are looking for ways to get a computer to replace all the humans and everyone thinks an LLM will solve all of their problems. But how do we ensure that we aren’t creating a catastrophic failure when we deploy LLMs in situations where we can’t afford to fail?

In this keynote, I will explore some of the issues we need to contend with when we put LLMs and other language technologies into an enterprise. I will touch on data preprocessing, governance, user trust and interpretation.

Bio

Kyla Quinn is the Technical Director of Data and Analytic Services Branch at the Australian Signals Directorate. In this role she provides strategic direction for staff involved in developing analytic tooling, from the AI and ML used in the back end through to user interfaces. Kyla has a background in engineering and linguistics and has recently submitted her PhD which is an evolutionary exploration of paradigm syncretism in the world’s languages through Bayesian analysis and LLM embeddings.

Panel Discussion

Information will be available soon.