Meeting link: You will receive meeting link after registration confirmation process.
Speaker: Henrique Lopes Cardoso, Associate Professor at the Faculty of Engineering of the University of Porto (FEUP), Senior researcher at the Artificial Intelligence and Computer Science Lab (LIACC), and Vice-President of the Portuguese Association for Artificial Intelligence (APPIA)
Abstract: Large language models are here to stay, and they have shown impressive performance in several benchmarks. Their integration with other tools, such as search engines and software development environments, allows for significantly improved productivity. Part of their success is related to scaling laws, which makes bigger language models exhibit emergent properties and surprising experimental results. At the same time, their capabilities seem to be very shallow regarding language understanding and common-sense reasoning, shedding light on their limitations and potential misuse. In this talk, while discussing these issues, we will understand how language models work and demystify the hype behind their potential as a path toward AGI. Prompt engineering is also discussed as the developing approach to tame large language models. We will also analyze current efforts to make smaller and more reliable models.
Short bio: Henrique Lopes Cardoso holds a PhD in Informatics Engineering from the University of Porto, is an Associate Professor at the Faculty of Engineering of the University of Porto (FEUP), a senior researcher at the Artificial Intelligence and Computer Science Lab (LIACC), and Vice-President of the Portuguese Association for Artificial Intelligence (APPIA). He teaches and researches on intelligent systems, with particular emphasis on multi-agent systems, natural language processing, and machine learning. He has participated in several national and international projects on these topics. He collaborates regularly with the European Commission as an expert in project evaluations. He is vice-chair of the working group on computational linguistics in the COST Action “Language in the Human-Machine Era (LITHME)”.