Uncertain⟨T⟩
13 days ago
- #programming
- #uncertainty
- #probabilistic-modeling
- People are attracted to confidence, even if it's misplaced, making nuance hard to brand.
- Software engineers often face uncertainty but must make binary decisions in code, which doesn't reflect real-world probabilistic scenarios.
- The Uncertain<T> type system encodes uncertainty directly into programming, allowing for probabilistic comparisons and operations.
- Monte Carlo sampling is used to model uncertainty by simulating random outcomes, like in slot machines.
- Uncertain<T> supports various probability distributions (normal, exponential, etc.) and statistical operations for real-world applications.
- Gradual migration to uncertain calculations can improve accuracy without requiring full rewrites.
- Sampling has computational costs, so balance between speed and precision is necessary.
- The goal is to acknowledge and handle uncertainty gracefully rather than pretending it doesn't exist.