Conversational AI, RAG, Multimodal Models & Teaching AI | Mandy Korpusik

What does it take to build conversational AI systems that are actually trustworthy? In this episode of AI Chronicles Season 1: AI Research, I speak with Professor Mandy Korpusik about the research challenges behind conversational AI, NLP, retrieval-augmented generation (RAG), multimodal models, and AI applications in health and nutrition. We also get into where semantic understanding still breaks down, whether RAG really solves hallucination, what changes when AI systems have to reason across both language and images, and how universities are adapting to a world where students increasingly want to jump straight into LLMs, AI agents, and coding assistants. Mandy also shares an interesting perspective on AI education: students still need strong foundations in machine learning and NLP, but they also need to learn how to use modern AI tools effectively once those fundamentals are in place. In this episode, we discuss: • Building trustworthy conversational AI systems • Hallucinations in high-stakes AI applications • Retrieval-Augmented Generation and its limitations • Semantic understanding in modern LLMs• Multimodal AI and vision-language models • AI applications in health and nutrition • Teaching machine learning and NLP in the age of AI agents • How students should learn to use AI coding assistants • The balance between AI fundamentals and modern tools If you're interested in LLM research, conversational AI, RAG, multimodal AI, NLP, or the future of AI education, this episode is for you.