Models Within Models - How Do LLMs Represent the World?

Explore a thought-provoking lecture by Harvard University Professor Martin Wattenberg examining how Large Language Models (LLMs) conceptualize and represent information about the world. Throughout this 75-minute presentation from UC Berkeley's EECS department, delve into the complex internal structures and mechanisms that enable LLMs to process, store, and generate knowledge. Learn about the hierarchical nature of these AI systems and how they develop internal representations of concepts, relationships, and real-world phenomena.