Csaba Szepesvári Says AI Doom Probabilities Should Be Intervals, Not Point Estimates
CsabaSzepesvari · x · 2026-10-11
Responding to a thread where someone quoted their P(doom) as 1e-5, RL researcher Csaba Szepesvári argues that giving a precise point estimate for AI extinction risk is meaningless—no one can be confident about a number like 1e-5, and the critique applies to everyone quoting such figures. He proposes that the community switch to reporting intervals (or more refined uncertainty statements) instead, which would make the debate more meaningful.
Related event: AI Researchers Debate P(doom): Some Claim Just 1-in-100,000(3 posts)→
More from AGI Musings
- Lawyer argues AI reports of context-window 'death' are strong evidence of experience — repligate · 2026-10-11
- "Your phone is a rock": tracing 3.3 million years of toolmaking from stone tools to pocket chips — kevrussell · 2026-10-11
- Prediction: humans will still be debating LLM consciousness 500 years from now — wordgrammer · 2026-10-11
- New essay 'Data At The Edge': the most valuable datasets are the ones we can't access — rebeccakaden · 2026-10-11
- iamtrask: AI can't master humor from public data alone — the punchline needs shared lived experience — iamtrask · 2026-10-11
- How Will Future Historians View Model Deprecation — When the Historians Are Models? — repligate · 2026-10-11