https://studiegids.vu.nl/en/courses/2024-2025/XB_0119Knowledge and understanding: at the end of the course, students will have a solid understanding of conversational AI concepts (e.g., dialogue acts), development and evaluation of dialogue systems, large language model (LLM) architectures and how LLMs can be deployed as conversational agents, challenges with LLMs as conversational agents and mitigation strategies, tool-augmented LLMs, grounding LLMs to external knowledge, learn the components of multi-agent and multi-modal agents, the role of LLMs in embodied environments, and understand considerations of pragmatics, storytelling, and efficiency. Applying knowledge and understanding: students will be able to implement conversational AI using LLMs and test their system on various conversational tasks. Making judgments: students will have a basic understanding of the ethical and societal implications of the developments in conversational AI. Communication skills: students will be able to write a scientific report about an original research question in collaboration with a group of students. Learning skills: students will be trained to develop conversational AI for multiple tasks and domains, develop an original and precise research question, and perform the corresponding experiments to answer their questions.Foundational concepts of conversational AI, such as dialogue acts. Concepts and architectures of Large Language models as components of conversational AI. Understanding model limitations and designing mitigation strategies to address these limitations. LLMs with tool augmentation and grounding to external knowledge. Design and development of multi-modal and multi-agent LLM systems. Adapt and build conversational AI for specific target domains and applications.Theoretical lectures and working group sessions (one of each per week)Multiple choice exams on theory (50%) and assignments (50%). Assessment details:Each of the three assignments must be at least 5/10.If a student scores below this grade, they can resubmit once based on feedback, a few days after the feedback is given by the instruction team.Resubmissions for only one of the three assignments are allowed for each student.The assignment grade must be at least 5.5/10 on average over the three assignments.The exam grade must be at least 5.5/10.If a student scores below this grade, they can take a resit.The literature list will be made available through CanvasStudents are expected to have solid programming skills in Python and general understanding of AI concepts.