Lecture 2
Technical Foundations & Lessons from the History of Governing Frontier Technology
Technical Foundations
This lecture provides a technical foundation for understanding how modern AI systems work. We'll cover the core concepts that underlie today's large-scale models, including stochastic gradient descent, neural networks, and key optimization techniques. Students will learn about model architectures used in foundation models, as well as common strategies for improving performance such as fine-tuning, quantization, and distillation. We'll examine how scale affects model capabilities, and introduce reasoning phenomena such as chain-of-thought reasoning. This technical grounding is essential for analyzing later governance challenges related to safety, evaluation, and access.