All About Biases in Artificial Intelligence
summary
Showcased are concrete harms and fixes: social media ads shift after changing gender, a chatbot turns racist in four hours, hiring models downrank CVs containing the word “woman,” and a consumer credit algorithm gives one spouse up to 20× more credit. Cited studies report women 73% more likely to be seriously injured and 177% more likely to die in crashes (female crash test dummy only commercially available 2023), 1.3M-patient research shows women 32% likelier to die with male surgeons, and criminal-risk models label Black defendants ~45% higher risk; proposals include explainable models and policy work (EU AI Act).
AI-written summary · about
chapters (16)
- 0:00Introduction
- 2:56Everyone has biases
- 4:44Historical biases in tech
- 6:51Gender bias quiz
- 10:26Gender data gap
- 15:27AI learns human bias
- 16:30AI bias in hiring & finance
- 18:34Generative AI stereotypes
- 20:21AI bias in justice systems
- 23:19Why bias correction matters
- 26:51Solutions against AI bias
- 28:16Q&A
- 31:29Regulation approaches
- 34:17Fixing human bias
- 37:34Can AI be unbiased?
- 40:28Progress and closing
related talks
34mAI and Gender: Building Fair AI Systems for the Future
22mGender Bias in Artificial Intelligence
56mAn AI with an Agenda: How Our Cognitive Biases Leak Into Machine Learning
21mFairness and bias in AI
38mSocial Implications of Bias in Machine Learning
26m(Un)ethical Artificial Intelligence: How to Keep the AI Fair for Everyone
37mThe AI 'Black Box'
30mWhy the Future of AI Is Feminist