In this interactive session, Arnoud Wils explained how large language models such as ChatGPT actually work. Students learned how words are converted into numerical representations, allowing computers to perform calculations with language. Using interactive applications, they could see how a computer processes text and how a GPT model generates sentences word by word.

We also explored what a large language model is, how it is trained, and what kind of data is used in that process. This included a discussion on ownership of training data and the ongoing legal debates between news media organizations and big tech companies.

By unpacking the mechanics behind generative AI, students gained insight into both its strengths and its limitations and risks.