Ethical Machines Predictions are Commands
4 snips
May 7, 2026 
Carissa Véliz, Associate Professor at Oxford and author of Prophecy, explores how predictions in AI function as power plays. She contrasts human forecasts with weather, warns that large-scale AI creates monocultures, and argues that claims of inevitability can become commands. The conversation urges questioning predictions, protecting autonomy, and planning for uncertain futures.
AI Snips
Chapters
Books
Transcript
Episode notes
Human Predictions Change The Future
- Predictions about people differ from predictions about things because people can respond and change behavior in light of the prediction.
- Carissa Véliz: clouds don't care about forecasts, but telling a student they'll fail will alter how you treat them and thus the outcome.
Talented Candidate Rejected By Algorithms
- Véliz recounts meeting a skilled person who was repeatedly desk rejected by hiring algorithms despite later getting high-paying executive offers.
- The algorithm favored the normal curve and missed outliers who were later proven highly competent.
Scale Turns One Bias Into Systemic Risk
- AI magnifies bias problems because similar models are trained on shared datasets and deployed broadly.
- Carissa Véliz: one algorithm can block a candidate across many companies, removing the diverse second chances humans provide.
