Overview: In this workshop, we will provide the audience with a general introduction to large language models (LLMs), which will include a brief history of their evolution from statistical to deep learning models as well as the technology underlying the current generation of LLMs. We will also present the audience with some current pitfalls of LLMs, including their tendency to hallucinate. Finally, we will have hands-on activities where we will compare the outputs of different LLMs (e.g. GPT-4 vs Claude) and highlight some of the pitfalls discussed in the workshop, including hallucinations as well as their sensitivity towards prompt manipulations.

Location: HYBRID. Online via Zoom, or in-person at Burnside Hall room 1104 (11th floor).
Instructor: Cesare Spinoso-Di Piano, PhD student in Computer Science, McGill University

IMPORTANT NOTE: for those attending the workshop online, you must log-into your McGill Zoom account. Otherwise, the link will not work. You can do so on the browser here: https://mcgill.zoom.us/ . For more information on McGill Zoom, please visit this website: https://www.mcgill.ca/tls/students/learning-resources/use-technology/learning-zoom