Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas 196 | Judea Pearl on Cause and Effect
18 snips
May 9, 2022 Judea Pearl, a trailblazer in causal inference and AI, shares his insights on the complexities of understanding causality. He delves into how we attribute credit or blame, emphasizing the need for a nuanced approach to cause and effect. Pearl discusses the significance of the 'do operator' in causal diagrams and its impact on AI and programming. He also explores the evolution of human curiosity and counterfactual thinking, linking it to cognitive advancements. The conversation highlights the essential relationship between causality, entropy, and our interpretations of data.
AI Snips
Chapters
Books
Transcript
Episode notes
Emergent Causality
- Causality, while seemingly intuitive, isn't fundamental in physics due to time-symmetric equations.
- Our perception of directionality emerges from factors like time, energy, and mass differences, like the sun and rooster example.
Counterfactual Calculus
- Counterfactuals are evaluated using causal diagrams, representing "listening" relationships between variables.
- This offers a compact way to represent a vast number of counterfactual scenarios, unlike resource-intensive philosophical approaches.
Barometer Example
- The barometer example illustrates how causal diagrams work. An arrow points from atmospheric pressure to barometer deflection, indicating a listening relationship.
- This clarifies that the barometer reacts to pressure, not vice-versa, a distinction robots without prior knowledge might miss.
