In-Class Discussions

Annotated outlines from class sessions

These annotated outlines capture the key topics, student contributions, and discussion threads from each class session. They serve as study aids and a record of our collective exploration of secure and trustworthy AI.

Societal Stakes and Ethical Frameworks

Week 2 Day 2

Wednesday, January 28, 2026

Worker displacement in tech and cybersecurity, the singularity debate and exponential AI progress, cognitive and skill degradation from AI over-reliance, and applying philosophical ethical frameworks to AI systems.

Core AI Values: Fairness

Week 3 Day 1

Monday, February 2, 2026

Insights from student questionnaires (six themes), current events (AI-only social media platforms), and a deep dive into fairness -- distributive, procedural, and interactional types, the Capuchin monkey experiment, and ML tasks where fairness matters.

Privacy, Autonomy, Safety, and Sustainability

Week 3 Day 2

Wednesday, February 4, 2026

AI-specific privacy threats (deepfake identity fraud, automated phishing), autonomy and sycophancy in AI systems, AI and human rights from cross-cultural perspectives, and frameworks for human-AI collaboration.

Student Case Study Presentations

Week 5 Day 1

Monday, February 16, 2026

Seven student-led case study presentations and discussions: NEDA Tessa chatbot, CBA worker displacement, SoftBank Pepper robot, Scatter Lab Luda chatbot, Adam Raine/ChatGPT, AI deepfake romance scams, and the AI-generated Pentagon explosion image. Thematic analysis of premature deployment vs. intentional misuse.

AI/ML Governance and Documentation

Week 5 Day 2

Wednesday, February 18, 2026

Governance etymology and cybernetics, the five-stage AI/ML pipeline: datasheets for datasets (Gebru et al.), model cards (Mitchell et al.) with live demos of Anthropic, Google, OpenAI, and Hugging Face platforms, international standards (ISO/IEC 42001, IEEE 7000), and the NIST AI Risk Management Framework.

AI Policy Current Events, Emergence, and Reinforcement Learning

Week 6 Day 2

Wednesday, February 25, 2026

Pre-class debate on the reported DoD/Anthropic Defense Production Act story, risks and ethics of AI in military applications, and the geopolitical context; midterm-exam logistics and guest-instructor announcements; recap of the 3Blue1Brown Transformer videos and emergence (BOIDs, swarm behavior); reinforcement learning, the AlphaGo documentary, and continual learning / catastrophic forgetting.

Reward Hacking and the University Education Model

Week 8 Day 1

Monday, March 9, 2026

Return-from-Spain check-in and recap of the prior week (instructor absent), upcoming midterm-exam logistics, and a discussion of natural human learning, reward hacking, and the purpose of a university education.

Midterm Review

Week 8 Day 2

Wednesday, March 11, 2026

In-class midterm review session: walkthrough of the study-guide topics and Q&A ahead of the comprehensive midterm exam covering material through Unit 6.

LLM-Assisted Development and a Concept-Drift Demo

Week 10 Day 1

Monday, March 30, 2026

Opening discussion of the "Rent a Human" platform; a video walkthrough of a five-phase LLM-assisted development cycle (Review → Brainstorm → Research → Plan → Execute) with a live demo building an interactive Streamlit app to visualize concept drift in a classifier under natural and adversarial conditions.

3Blue1Brown — Transformers (Deep Learning Ch. 5)

Video Outline Third-Party

Companion outline to Grant Sanderson’s 3Blue1Brown video

A section-by-section outline of the 3Blue1Brown video introducing the Transformer architecture: word embeddings, the high-level data flow through a transformer, the deep-learning premise, unembedding, and softmax with temperature. Summarizes third-party material — see the original video for full context.

3Blue1Brown — Attention in Transformers (Deep Learning Ch. 6)

Video Outline Third-Party

Companion outline to Grant Sanderson’s 3Blue1Brown video

A step-by-step outline of the 3Blue1Brown video on attention: queries and keys, the attention pattern, masking, context size, value matrices, multi-headed attention, and why attention works. Summarizes third-party material — see the original video for full context.