Beyond P(doom) for AI Risk: Quantifying Uncertainty Without Probability
Beyond P(doom) for AI Risk: Quantifying Uncertainty Without Probability
Publish Date: 2026-05-06 13:23:00
Source Domain: cset.georgetown.edu
- The article highlights the new risks introduced by artificial intelligence, which could be catastrophic or even existential in nature.
- Existing data and theory for assessing AI risks are insufficient; policymakers often rely on expert guesses.
- Traditional probability assessments may not be the best tools for understanding AI risks due to their inherent uncertainties.
- The report introduces Belief and Plausibility as an alternative, mathematically rigorous framework.
- This new approach uses familiar vocabulary and reduces the assessment process to just two simple questions for policymakers.
