MECD Talk

Orador: Hugo Oliveira

Sumário da palestra:

Language Models have been here for decades, but only recently had an impact way beyond the domains of Natural Language Processing (NLP) and Artificial Intelligence.
Current models, like GPT-4 and LLaMA-2, take advantage both of recent advances in Machine Learning and of the easy access to huge volumes of data and they summarise, translate, and answer questions, among others, without having been training explicitly in such tasks.
These models have had much attention, especially after the release of ChatGPT, whose potential has been discussed not only in the academic and business worlds, but also by the general public.
In this presentation, Large Language Models (LLMs) will be contextualised in the scope of Natural Language Processing and some of their limitations will be highlighted.

Biografia:

Hugo Gonçalo Oliveira is an Associate Professor at the Department of Informatics Engineering of the University of Coimbra and a researcher of the Cognitive and Media Systems (CMS) group of the Center of Informatics and Systems of the same university (CISUC). His main research areas are Artificial Intelligence, specifically, Natural Language Processing (NLP) and Computational Creativity (CC). Besides his contribution to the creation of multiple computational linguistic resources for Portuguese, he has participated in several National and European projects, including the coordination of the FCT pilot AIA (2018-20) and two P2020 projects in co-promotion, Flowance (2020-23) and SmartEDU (2021-23). Hugo is the author of circa 150 papers in peer-reviewed journals and conference proceedings. He is a regular member of the program committee of the main events in NLP and CC, and is the executive editor of the journal Linguamática.
His PhD thesis was awarded the best in the Computational Processing of Portuguese (2011-2014) and he had a special recognition by the World Cultural Council (2022), in the category Education.