Tue, Sep 22
Topic 2. How Computers Work
Inside a Neural Network: Weights, Layers, and Learning
Please complete the tasks and readings listed below before class on Tue, Sep 22.
Required Readings
Reading
Crawford, Kate. “Atlas of AI.” Sydney Ideas, University of Sydney, 2021. Link
Ideally, watch the whole thing. If you're short on time, watch minutes 11-24. Discusses training data, ImageNet, classification, and how human categories shape what AI systems learn to recognize.
Reading
Sanderson, Grant. “But What Is a Neural Network?” 3Blue1Brown, 5 Oct. 2017. Link
Watch approximately the first 8–9 minutes for an intuitive introduction to neurons, layers, and learned features.
Optional Readings
Reading
Google. Teachable Machine [Interactive]. Link
Try object classification; test unfamiliar backgrounds and examples.
Guiding Questions
- We’ve been learning from data for a long time (i.e., that what statistics does)! So what are people talking about when they say, “nobody knows how the model works?”
- How are Deep Neural Networks similar and different from what came before, and what new benefits / harms, possibilites / risks do they raise?
In This Class
Societal / Ethical Questions
- When is a system too opaque to trust?
- Is higher accuracy worth lower interpretability?
Please complete the tasks listed below tonight on Tue, Sep 22.
Submit to Canvas
Career Module
Career Module 2: Three Possible Lives and SMART GoalsDue Tu, Sep 22 at 11:59 PM
Before next class
| Class prep | 2 readings to complete | View |