Large Language Models (LLMs) and the Changing Face of Computational Linguistics

Abstract: Computational Linguistics has gone through two major conceptual shifts: from symbolic to machine learning approaches, and the Deep Learning revolution.  The surprising ability of Transformers and other Deep Learning models to directly capture sentence and paragraph-length contexts – coupled with enormous training data and compute – enable them to supplant conventional NLP technologies for many practical applications.  In the near future LLMs will realise the slogans “Natural Language is the New Programming Language” and “Natural Language is the New Knowledge Representation Language”.  This talk discusses the implications of this revolution for related fields, and language learning more generally.