Generative AI: Technical and Social

The development of Generative AI is rapidly advancing beyond simple next-token prediction, evolving into sophisticated intelligent systems known as AI agents. These agents can manage databases and tools, perform reasoning and reflection, retain historical context, and iteratively plan and execute actions. The potential of AI agents includes: i) delivering unprecedented levels of intelligence and workflow automation (e.g., vibe coding) across nearly all areas of life; ii) simulating human behavior to enable faster testing, evaluation, and improvement; and iii) driving new social, economic, and market innovations. In this class, we will take a “learning by doing” approach to explore AI agents. We will reproduce results from state-of-the-art research papers, gain hands-on experience with recent AI agent applications, and implement simplified versions of these systems from scratch. Along the way, we will discuss both the technical and social opportunities and challenges raised by these applications. This course will be coding-intensive, but we’ll also leverage coding agents (e.g., Cursor AI agent) to support our work.