NLP is Dead, Long Live NLP

This event will be held on Zoom

Large language models with multi-tasking capabilities have emerged in the last several years and made traditional NLP tasks, such as ASR, machine translation, summarization, text generation, sentiment analysis, as well as intermedia NLP tasks such as POS tagging, parsing, named entity recognition, textual entailment, etc.,  obsolete. It is fair to say that the field of NLP, as defined by these tasks, is dead. It is also evident in conferences such as  ACL and EMNLP where the vast majority of work is based on some form of LLM pre-training or post-training. Indeed, LLM research has both expanded beyond traditional NLP areas and replaced the latter. Quo Vadis NLP?

In this talk, I argue that, following LLMs, language research is more relevant than ever. The emergence of agentic systems marks a new milestone in human-machine interactions.  I propose a new framework of language research that looks into mental world modeling, which parallels physical world modeling. Language both represents and shapes the “mental world models” of humans, enabling us to reason, plan, emote, learn, innovate and create. Natural language research, therefore, should endeavor to come up with better and more efficient mental world models than current LLMs. The scope of “language” should expand to include non-iconographic languages such as gesture and music.  NLP research can thrive again, leading to new innovations in AI.

Bio:
Pascale Fung is a Senior Director of AI Research at Meta. She is also a Chair Professor of Electronic & Computer Engineering at The Hong Kong University of Science & Technology (HKUST), and a visiting professor at the Central Academy of Fine Arts in Beijing. She is an elected Fellow of the Association for Advancement of AI (AAAI), IEEE, ACL and ISCA.  She served on the Board of Governors of the IEEE Signal Processing Society. She served as Editor-in-Chief and Associate Editor for all the top academic journals in the field of speech and language processing. Her team has won several best and outstanding paper awards at ACL, ACL and NeurIPS workshops.  

She is an expert on the Global Future Council of the World Economic Forum and is on the Expert Network of the UN Advisory Body  on AI. She is a  member of the Partnership on AI to Benefit People and Society. She has been a keynote or invited speaker in 250 conferences and panels to speak to both the AI technical community and the public. She has been invited as an expert on AI governance by the United Nations, EU, EC, European Parliament, the UAE, Saudi Arabia, Japan, Singapore, the US Academics of Sciences, Medicine, and Engineering, among others. She was invited to a private audience with Pope Francis in 2023 to discuss AI ethics. She is listed as one of the Forbes Asia 2024. A speaker of seven European and Asian languages, she is currently based in Meta-FAIR in Paris.