Keynote Talk: Can Large Language Models Make Decisions? Mathematical Formulations and Open Problems

Large Language Models (LLMs) and foundation models are increasingly being considered as agents for decision-making in domains ranging from autonomous systems to business processes. This talk will explore mathematical frameworks that applied probabilists, optimizers, and operations researchers can use to analyze and design such systems. We will discuss some recent results and present open problems formulated in the language of optimization, Markov Decision Processes, filtering, and control, highlighting opportunities for future research.