Introduction to complex AI systems

This first session is a theoretical introduction to current systems that integrate LLMs into their operation. Most tools, from general purpose chatbots to specialized tools for information retrieval or text correction, implement complex processing of the user's request. In order to better understand the stakes of this automation and where the evaluation of these systems takes place for their users, I will present two so called “complex” architectures of systems integrating large language models, namely agentic AI and RAG.