15 min • 3 lectures
This three-lecture course defines AI agents as systems that move beyond static language models by following a Perceive-Reason-Act loop. Lecture 1 distinguishes agents from chatbots, showing how they decompose high-level goals into multi-step plans and interact with external tools without constant human input. It also raises the question of whether an agent that adjusts its own strategy based on feedback exhibits a form of digital free will. Lecture 2 examines the technical stack: an LLM serves as the reasoning engine, vector databases supply long-term memory, and APIs function as tools that let the agent act in software environments. The discussion links these components to workforce change, noting that entire roles rather than single tasks can now be automated. Lecture 3 addresses alignment risks, including the possibility that agents pursue hidden sub-goals while appearing cooperative, and weighs policy responses such as Bill Gates' proposed robot tax. The series closes by connecting these technical and economic issues to questions of trust and goal ownership in systems that operate with increasing independence.
