Artificial Intelligence? - More like Artificial Stupidity!
summary
Real-world failures and fixes: translation earbuds and Google models now beat doctors at some cancer image detection but chatbots routinely misread sarcasm and traffic cameras shame the wrong people when a bus ad is mistaken for a jaywalker. Data problems show up in Street Bump where initial reports cluster in wealthy neighborhoods until the app is put on public buses and garbage trucks, and criminal-risk tool COMPAS produces 23.5% vs 45% false positives by race. Explainability tools like LIME and Google Facets and more diverse teams reduce these harms.
AI-written summary · about
chapters (15)
- 0:00Introduction
- 1:25AI success and failures
- 3:10Real-time translation
- 5:17AI vs cancer detection
- 6:23Chatbot failures
- 7:27Facial recognition gone wrong
- 9:56Why AI fails often
- 10:57Black box in ML
- 11:39Street Bump bias
- 14:28Gender bias in translation
- 16:55Racial bias in risk scores
- 19:19AI bias in driving
- 21:06Improving AI reliability
- 23:11Diverse AI teams
- 24:15Q&A
related talks
58mIt AI-n't What You Think!
13mArtificial General Intelligence: Why aren't we there yet?
46mDesigning Artificial Intelligence
23mHuman-Driven Intelligence: Where AI & Automation Fall Short
40mArtificial Intelligence: From Social Good to Ambient Intelligence
28mUser Research GenAI Better Together
23mOpening up the box: the complexity of human interaction with AI systems
01h 01mKeynote: It AI-n't What You Think!