Bad computers, bad people
summary
Shows concrete failures: Microsoft’s chatbot Tay learned hate speech within hours and had to be shut down, Google’s image model labeled Black people as gorillas until retrained, YouTube’s restricted mode filtered innocuous LGBTQ videos, and consumer VR induces motion sickness disproportionately in women; she cites a study that women were 47% more likely to suffer serious or fatal injuries because crash testing used male dummies. Argues algorithmic bias comes from training data, design choices and lack of diverse teams.
AI-written summary · about
chapters (23)
- 0:00Intro
- 7:00Computer Science 101
- 14:28Microsoft Chat Bot
- 15:20Tabak
- 16:04Google Photos
- 16:45Google Image Recognition
- 17:38YouTube Restricted Mode
- 18:41VR
- 19:05Motion Sickness
- 19:50Cute Baby
- 21:16Is an algorithm racist
- 22:00Is an algorithm sexist
- 22:29Biases in algorithms
- 23:24Check yourself
- 24:36Privilege
- 25:30Charging users
- 25:56College ID
- 26:20Work History
- 27:11Programming Bias
- 28:13Diversity
- 32:03Bad computers
- 32:36Asymptote
- 33:06Question for Terri
related talks
29mUnintended consequences of technology: Enabling NextGen discrimination
38mSocial Implications of Bias in Machine Learning
20mThere is no such thing as bias-free tech. How models shape our world.
35mDesigning Inclusive and Accessible Experiences in the Era of AI
34m(Un)Ethical Machines? Why AI Bias Is a Problem and What to Do About It
37mThe AI 'Black Box'
15mFrom Personalized to Programmed
28mBreaking the Chains of Oppressive Software