Women and Non-Binary in PyTorch Breakfast with Quiana Berry, Red Hat
summary
Covers concrete harms and fixes: cites 95% non-consensual deepfakes and $78B misinfo cost (2020), warns AI energy use could match the Netherlands by 2027 and that women are 22% of the AI workforce; shows GPT‑4 safety tuning reduces disallowed responses by 82% but can still be jailbreaked; gives cases of wrongful ID arrests, exploited Kenyan data labelers, and wins like lung‑cancer and diabetic‑retinopathy detection, flood forecasting and anti‑trafficking work; recommends ethics committees, recruiting underrepresented groups, and binding governance (EU AI Act, Linux Foundation principles).
AI-written summary · about
chapters (9)
- 0:00Welcome
- 3:10Ethical AI and deep fakes
- 10:15Defining ethical AI
- 14:09Bias in AI systems
- 18:01Enterprise bias and ML pipeline
- 22:32Harms of biased AI
- 26:21Researchers' warnings and AI governance
- 33:22AI benefits versus regulation
- 38:01Poem and closing reflections
related talks
30mFireside chat: Breaking the wall of unethical AI
44mPanel: AI Feminism
47mHow gender imbalance is holding back the future of AI
50mIs ChatGPT (and Other AI) the Enemy of Diversity?
29mUnintended consequences of technology: Enabling NextGen discrimination
40mResponsible AI Practices: Fairness in ML
45mThe Myth of Neutrality: How AI is widening social divides
55mPanel: Gender bias in AI - In collaboration with Women in Data Science Belgrade