Big Technology Podcast Are We Too Obsessed With AI Predictions? — With Carissa Véliz
96 snips
Apr 22, 2026 
Carissa Véliz, an Oxford philosopher and author focused on privacy and tech ethics, dives into why AI prediction can become a tool of power. She explores hiring and lending algorithms, surveillance and protest anonymity, prediction markets, generative AI’s taste for plausibility, and why humor and art can push back against a forecast-driven world.
AI Snips
Chapters
Books
Transcript
Episode notes
Black Box Lending Replaces Reasons With Guesses
- Véliz says predictive lending is harder to contest than rule-based lending because predictions are guesses, not facts.
- A bank can state "you need $10,000" and be checked, but a black-box denial gives applicants no clear error or path to improve.
You Cannot Audit Missing Counterfactual Lives
- Even audited predictive systems remain problematic because denied people never generate the counterfactual data that could prove the model wrong.
- Véliz says this creates Kafkaesque systems where people cannot learn the rules and start guessing what the algorithm "wants."
AI Prediction Echoes Ancient Oracles
- Véliz compares today's faith in AI prediction to ancient trust in the Oracle of Delphi and astrology as elite decision tools.
- She distinguishes predicting physical systems like weather or floods from predicting social behavior, where feedback loops and ambiguity distort outcomes.
