Artificial Intelligence? - more like Artificial Stupidity!
summary
Shows real wins like Google Translate earbuds and ML models that outperform doctors on cancer scans, then shows failures: a camera misreads a printed face on a bus as a jaywalker, a pothole app skewed to wealthy neighborhoods until buses fixed coverage, Tay becomes racist within a day, and risk tool findings (notably COMPAS: 45% vs 23% false positives on Black vs white defendants) reveal biased outcomes. Presents practical fixes: dataset auditing, LIME explainability and tools like Google Facets and the Data Ethics Canvas.
AI-written summary · about
chapters (5)
- 0:00Start
- 0:12Artificial Intelligence?
- 45:15More diverse teams
- 46:41Remove Data
- 49:13What are your questions?
related talks
41mWhen AI Failures Reflect Our Own Social Biases
30mHow to Lie with AI: Understanding Bias, Ethics, and the Hidden Risks in ML
54mAI: Actual Intelligence
26m(Un)ethical Artificial Intelligence: How to Keep the AI Fair for Everyone
36mKeynote: Ethics in AI: Building Fair and Just Intelligence Systems
39mThe Ethics of Artificial Intelligence - What...
49mAn AI with an Agenda
46mAI vs Humans Rewriting the Rules of Success