Large language models have taken over Norwegian media companies, and the technology is being used in increasingly more phases of news production, from idea development, transcription, translation, to summarization.

But what exactly is a language model, and how do they work? Samia Touileb is an associate professor in language technology at the Department of Information and Media Studies at the University of Bergen, works in SFI MediaFutures and is an expert on language models. In this session, she will “open the hood” and show us what lies inside the large language models.

What is so-called deep learning? What architectures are hidden between our input and the model’s output? Where are the language models’ training data sourced from? And what ethical blind spots and problematic biases exist in this data?

The session is part of the Nordic Media Days and required a conference ticket.