Explore the ethical dimensions of building ML-powered products in this comprehensive lecture covering past and present tech ethics, current and near-future ML ethics, and future AI ethics. Delve into key themes such as alignment, trade-offs, and humility while examining the tech industry's ethical crisis. Learn about tracking carbon emissions, deceptive design patterns, and the increased salience of ethics in machine learning. Investigate crucial questions surrounding model fairness, representation in ML, system accountability, data ownership, and the ethical considerations of whether certain technologies should be built at all. Discover how medicine is leading the way in responsible ML implementation and explore frontiers like AI rights and existential risks. Gain insights into practical steps for addressing ethical concerns and building ML systems that are both technically sound and socially responsible.
Overview
Syllabus
Overview and Context
Themes: Alignment, Trade-Offs, Humility
The Tech Industry's Ethical Crisis
Tracking carbon emissions
Deceptive design and dark patterns
Why is ethics more salient in ML?
Is the model fair?
Representation and ML
Is the system accountable?
Who owns the data?
Should this be built at all?
Medicine leads the way on responsible ML
AI Snake Oil and AI Ethics
Frontiers: AI Rights and x-risk
What is to be done?
Coda: building ML well and building ML for good
Taught by
The Full Stack
