Iris Talk: Entendendo o ChatGPT

In this talk, I introduce Large Language Models (LLMs), explaining how they work, how they learn, and their role in modern artificial intelligence. I briefly cover the history of conversational agents and the key ideas behind the Transformer architecture, with an emphasis on attention mechanisms, and discuss how these models are trained at scale and applied to tasks such as conversational AI and text generation.

The talk was presented in Portuguese at the Institute of Computing of Unicamp.