Tue, Sep 29
Topic 2. How Computers Work
Language Models: How can predicting the next word produce something so useful?
Please complete the tasks and readings listed below before class on Tue, Sep 29.
Required Readings
Optional Readings
Reading
Welch, S. (2025, July 25). But how do AI images and videos actually work? Guest video by Welch Labs. 3Blue1Brown. Link
A more technical video about how generative images and videos are produced.
Reading
Doctorow, C. (2026, September 16). How an AI moratorium can save AI bosses. Pluralistic. Link
Reading
Klein, E. (2026, September 15). Ezra Klein podcast: Matt Sheehan. The New York Times. Link
Reading
Zewe, Adam. “Study: Transparency Is Often Lacking in Datasets Used to Train Large Language Models.” MIT News, 30 Aug. 2024. Link
Read the article.
Topic / Focus
Large language models learn statistical patterns in enormous amounts of text and use those patterns to predict what comes next. By looking inside the technical process – and at the larger systems built around LLMs – we can better evaluate claims about what AI systems “know,” “decide,” or “want.”
Guiding Questions
- How does an LLM learn to predict what comes next?
- How do embeddings, transformers, and attention represent language and context?
- How does a language model become a chatbot?
- What roles do training data, human feedback, system instructions, and other software components play in producing a response?
In This Class
Societal / Ethical Questions
- How does the language we use to describe AI shape how we understand its capabilities?
- Where, if anywhere, does “agency” exist in an AI system?
- Who decides what counts as desirable chatbot behavior?
- Does focusing on powerful or autonomous “AI” distract us from more immediate risks and human design choices?
Please complete the tasks listed below tonight on Tue, Sep 29.
Submit to Canvas
Homework
Teachable Machine (Supervised Learning) ReflectionDue Tu, Sep 29 at 11:59 PM